pylearn: dataset.py annotate

annotate dataset.py @ 320:b1da46b9b901

bug fixed when initializing NArraysDataSet with 1-d arrays. This now raises an Exception

author	Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
date	Wed, 11 Jun 2008 17:11:37 -0400
parents	6441e568745e
children	f03ae06fadc8

rev	line source
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	1
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	2 from lookup_list import LookupList as Example
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	3 from misc import unique_elements_list_intersection
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	4 from string import join
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	5 from sys import maxint
171 895b4b60f5e8 bugfix. Otherwise the example was writed over and not a new one was returned Frederic Bastien <bastienf@iro.umontreal.ca> parents: 167 diff changeset	6 import numpy, copy
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	7
166 ee11ed427ba8 Created exceptions.py Joseph Turian <turian@gmail.com> parents: 163 diff changeset	8 from exceptions import *
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	9
110 8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	10 class AttributesHolder(object):
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	11 def __init__(self): pass
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	12
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	13 def attributeNames(self):
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	14 raise AbstractFunction()
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	15
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	16 def setAttributes(self,attribute_names,attribute_values,make_copies=False):
134 3f4e5c9bdc5e Fixes to ApplyFunctionDataSet and other things to make learner and mlp work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 132 diff changeset	17 """
3f4e5c9bdc5e Fixes to ApplyFunctionDataSet and other things to make learner and mlp work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 132 diff changeset	18 Allow the attribute_values to not be a list (but a single value) if the attribute_names is of length 1.
3f4e5c9bdc5e Fixes to ApplyFunctionDataSet and other things to make learner and mlp work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 132 diff changeset	19 """
3f4e5c9bdc5e Fixes to ApplyFunctionDataSet and other things to make learner and mlp work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 132 diff changeset	20 if len(attribute_names)==1 and not (isinstance(attribute_values,list) or isinstance(attribute_values,tuple) ):
3f4e5c9bdc5e Fixes to ApplyFunctionDataSet and other things to make learner and mlp work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 132 diff changeset	21 attribute_values = [attribute_values]
110 8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	22 if make_copies:
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	23 for name,value in zip(attribute_names,attribute_values):
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	24 self.__setattr__(name,copy.deepcopy(value))
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	25 else:
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	26 for name,value in zip(attribute_names,attribute_values):
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	27 self.__setattr__(name,value)
193 cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	28
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	29 def getAttributes(self,attribute_names=None, return_copy=False):
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	30 """
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	31 Return all (if attribute_names=None, in the order of attributeNames()) or a specified subset of attributes.
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	32 """
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	33 if attribute_names is None:
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	34 attribute_names = self.attributeNames()
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	35 if return_copy:
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	36 return [copy.copy(self.__getattribute__(name)) for name in attribute_names]
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	37 else:
cb6b945acf5a Complete redesign of learner... Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 188 diff changeset	38 return [self.__getattribute__(name) for name in attribute_names]
110 8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	39
8fa1ef2411a0 Worked on OneShotTLearner and implementation of LinearRegression bengioy@bengiomac.local parents: 105 diff changeset	40 class DataSet(AttributesHolder):
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	41 """A virtual base class for datasets.
813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	42
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	43 A DataSet can be seen as a generalization of a matrix, meant to be used in conjunction
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	44 with learning algorithms (for training and testing them): rows/records are called examples, and
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	45 columns/attributes are called fields. The field value for a particular example can be an arbitrary
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	46 python object, which depends on the particular dataset.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	47
241 ddb88a8e9fd2 If I understand properly, the length of an unbounded stream is sys.maxint delallea@opale.iro.umontreal.ca parents: 231 diff changeset	48 We call a DataSet a 'stream' when its length is unbounded (in which case its __len__ method
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	49 should return sys.maxint).
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	50
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	51 A DataSet is a generator of iterators; these iterators can run through the
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	52 examples or the fields in a variety of ways. A DataSet need not necessarily have a finite
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	53 or known length, so this class can be used to interface to a 'stream' which
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	54 feeds on-line learning (however, as noted below, some operations are not
241 ddb88a8e9fd2 If I understand properly, the length of an unbounded stream is sys.maxint delallea@opale.iro.umontreal.ca parents: 231 diff changeset	55 feasible or not recommended on streams).
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	56
813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	57 To iterate over examples, there are several possibilities:
90 a289b8bed64c corrected comment Frederic Bastien <bastienf@iro.umontreal.ca> parents: 88 diff changeset	58 - for example in dataset:
a289b8bed64c corrected comment Frederic Bastien <bastienf@iro.umontreal.ca> parents: 88 diff changeset	59 - for val1,val2,... in dataset:
a289b8bed64c corrected comment Frederic Bastien <bastienf@iro.umontreal.ca> parents: 88 diff changeset	60 - for example in dataset(field1, field2,field3, ...):
a289b8bed64c corrected comment Frederic Bastien <bastienf@iro.umontreal.ca> parents: 88 diff changeset	61 - for val1,val2,val3 in dataset(field1, field2,field3):
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	62 - for minibatch in dataset.minibatches([field1, field2, ...],minibatch_size=N):
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	63 - for mini1,mini2,mini3 in dataset.minibatches([field1, field2, field3], minibatch_size=N):
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	64 Each of these is documented below. All of these iterators are expected
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	65 to provide, in addition to the usual 'next()' method, a 'next_index()' method
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	66 which returns a non-negative integer pointing to the position of the next
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	67 example that will be returned by 'next()' (or of the first example in the
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	68 next minibatch returned). This is important because these iterators
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	69 can wrap around the dataset in order to do multiple passes through it,
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	70 in possibly unregular ways if the minibatch size is not a divisor of the
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	71 dataset length.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	72
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	73 To iterate over fields, one can do
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	74 - for field in dataset.fields():
46 c5b07e87b0cb comments modif made by Yoshua Frederic Bastien <bastienf@iro.umontreal.ca> parents: 45 diff changeset	75 for field_value in field: # iterate over the values associated to that field for all the dataset examples
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	76 - for field in dataset(field1,field2,...).fields() to select a subset of fields
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	77 - for field in dataset.fields(field1,field2,...) to select a subset of fields
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	78 and each of these fields is iterable over the examples:
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	79 - for field_examples in dataset.fields():
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	80 for example_value in field_examples:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	81 ...
241 ddb88a8e9fd2 If I understand properly, the length of an unbounded stream is sys.maxint delallea@opale.iro.umontreal.ca parents: 231 diff changeset	82 but when the dataset is a stream (unbounded length), it is not recommended to do
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	83 such things because the underlying dataset may refuse to access the different fields in
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	84 an unsynchronized ways. Hence the fields() method is illegal for streams, by default.
132 f6505ec32dc3 Updated documentation slightly Joseph Turian <turian@gmail.com> parents: 128 diff changeset	85 The result of fields() is a L{DataSetFields} object, which iterates over fields,
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	86 and whose elements are iterable over examples. A DataSetFields object can
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	87 be turned back into a DataSet with its examples() method::
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	88 dataset2 = dataset1.fields().examples()
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	89 and dataset2 should behave exactly like dataset1 (in fact by default dataset2==dataset1).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	90
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	91 Note: Fields are not mutually exclusive, i.e. two fields can overlap in their actual content.
813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	92
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	93 Note: The content of a field can be of any type. Field values can also be 'missing'
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	94 (e.g. to handle semi-supervised learning), and in the case of numeric (numpy array)
46 c5b07e87b0cb comments modif made by Yoshua Frederic Bastien <bastienf@iro.umontreal.ca> parents: 45 diff changeset	95 fields (i.e. an ArrayFieldsDataSet), NaN plays the role of a missing value.
c5b07e87b0cb comments modif made by Yoshua Frederic Bastien <bastienf@iro.umontreal.ca> parents: 45 diff changeset	96 What about non-numeric values? None.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	97
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	98 Dataset elements can be indexed and sub-datasets (with a subset
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	99 of examples) can be extracted. These operations are not supported
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	100 by default in the case of streams.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	101
317 14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	102 - dataset[:n] returns an Example with the n first examples.
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	103
317 14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	104 - dataset[i1:i2:s] returns an Example with the examples i1,i1+s,...i2-s.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	105
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	106 - dataset[i] returns an Example.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	107
317 14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	108 - dataset[[i1,i2,...in]] returns a. Example with examples i1,i2,...in.
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	109
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	110 A similar command gives you a DataSet instead of Examples :
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	111
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	112 - dataset.subset[:n] returns a DataSet with the n first examples.
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	113
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	114 - dataset.subset[i1:i2:s] returns a DataSet with the examples i1,i1+s,...i2-s.
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	115
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	116 - dataset.subset[i] returns a DataSet.
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	117
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	118 - dataset.subset[[i1,i2,...in]] returns a DataSet with examples i1,i2,...in.
14081904d8f3 doc updated regarding __getitem__ returning LookupList and .subset returning a DataSet Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 316 diff changeset	119
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	120
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	121 - dataset.<property> returns the value of a property associated with
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	122 the name <property>. The following properties should be supported:
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	123 - 'description': a textual description or name for the dataset
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	124 - 'fieldtypes': a list of types (one per field)
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	125 A DataSet may have other attributes that it makes visible to other objects. These are
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	126 used to store information that is not example-wise but global to the dataset.
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	127 The list of names of these attributes is given by the attribute_names() method.
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	128
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	129 Datasets can be concatenated either vertically (increasing the length) or
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	130 horizontally (augmenting the set of fields), if they are compatible, using
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	131 the following operations (with the same basic semantics as numpy.hstack
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	132 and numpy.vstack):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	133
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	134 - dataset1 \| dataset2 \| dataset3 == dataset.hstack([dataset1,dataset2,dataset3])
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	135
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	136 creates a new dataset whose list of fields is the concatenation of the list of
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	137 fields of the argument datasets. This only works if they all have the same length.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	138
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	139 - dataset1 & dataset2 & dataset3 == dataset.vstack([dataset1,dataset2,dataset3])
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	140
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	141 creates a new dataset that concatenates the examples from the argument datasets
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	142 (and whose length is the sum of the length of the argument datasets). This only
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	143 works if they all have the same fields.
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	144
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	145 According to the same logic, and viewing a DataSetFields object associated to
46 c5b07e87b0cb comments modif made by Yoshua Frederic Bastien <bastienf@iro.umontreal.ca> parents: 45 diff changeset	146 a DataSet as a kind of transpose of it, fields1 & fields2 concatenates fields of
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	147 a DataSetFields fields1 and fields2, and fields1 \| fields2 concatenates their
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	148 examples.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	149
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	150 A dataset can hold arbitrary key-value pairs that may be used to access meta-data
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	151 or other properties of the dataset or associated with the dataset or the result
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	152 of a computation stored in a dataset. These can be accessed through the [key] syntax
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	153 when key is a string (or more specifically, neither an integer, a slice, nor a list).
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	154
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	155 A DataSet sub-class should always redefine the following methods:
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	156 - __len__ if it is not a stream
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	157 - fieldNames
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	158 - minibatches_nowrap (called by DataSet.minibatches())
269 fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	159 For efficiency of implementation, a sub-class might also want to redefine
72 2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	160 - valuesHStack
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	161 - valuesVStack
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	162 - hasFields
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	163 - __getitem__ may not be feasible with some streams
2b6656b2ef52 Changed docs slightly Joseph Turian <turian@iro.umontreal.ca> parents: 66 diff changeset	164 - __iter__
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	165 A sub-class should also append attributes to self._attribute_names
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	166 (the default value returned by attributeNames()).
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	167 By convention, attributes not in attributeNames() should have a name
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	168 starting with an underscore.
3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	169 @todo enforce/test that convention!
2 3fddb1c8f955 Rewrote DataSet interface and created FiniteDataSet interface. bengioy@bengiomac.local parents: 1 diff changeset	170 """
1 2cd82666b9a7 Added statscollector and started writing dataset and learner. bengioy@esprit.iro.umontreal.ca parents: 0 diff changeset	171
83 c0f211213a58 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 82 diff changeset	172 numpy_vstack = lambda fieldname,values: numpy.vstack(values)
c0f211213a58 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 82 diff changeset	173 numpy_hstack = lambda fieldnames,values: numpy.hstack(values)
77 1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	174
292 174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	175 def __init__(self, description=None, fieldnames=None, fieldtypes=None):
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	176 """
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	177 @type fieldnames: list of strings
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	178 @type fieldtypes: list of python types, same length as fieldnames
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	179 @type description: string
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	180 @param description: description/name for this dataset
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	181 """
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	182 def default_desc():
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	183 return type(self).__name__ \
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	184 + " ( " + join([x.__name__ for x in type(self).__bases__]) + " )"
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	185
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	186 #self.fieldnames = fieldnames
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	187
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	188 self.fieldtypes = fieldtypes if fieldtypes is not None \
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	189 else [None]*1 #len(fieldnames)
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	190
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	191 self.description = default_desc() if description is None \
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	192 else description
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	193 self._attribute_names = ["description"]
292 174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	194
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	195 attributeNames = property(lambda self: copy.copy(self._attribute_names))
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	196
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	197 def __contains__(self, fieldname):
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	198 return (fieldname in self.fieldNames()) \
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	199 or (fieldname in self.attributeNames())
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	200
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	201 def __iter__(self):
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	202 """Supports the syntax "for i in dataset: ..."
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	203
292 174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	204 Using this syntax, "i" will be an Example instance (or equivalent) with
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	205 all the fields of DataSet self. Every field of "i" will give access to
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	206 a field of a single example. Fields should be accessible via
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	207 i["fielname"] or i[3] (in the order defined by the elements of the
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	208 Example returned by this iterator), but the derived class is free
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	209 to accept any type of identifier, and add extra functionality to the iterator.
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	210
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	211 The default implementation calls the minibatches iterator and extracts the first example of each field.
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	212 """
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	213 return DataSet.MinibatchToSingleExampleIterator(self.minibatches(None, minibatch_size = 1))
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	214
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	215 def __len__(self):
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	216 """
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	217 len(dataset) returns the number of examples in the dataset.
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	218 By default, a DataSet is a 'stream', i.e. it has an unbounded length (sys.maxint).
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	219 Sub-classes which implement finite-length datasets should redefine this method.
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	220 Some methods only make sense for finite-length datasets.
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	221 """
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	222 return None
174374d59405 merge James Bergstra <bergstrj@iro.umontreal.ca> parents: 290 diff changeset	223
78 3499918faa9d In the middle of designing TLearner bengioy@bengiomac.local parents: 77 diff changeset	224
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	225 class MinibatchToSingleExampleIterator(object):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	226 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	227 Converts the result of minibatch iterator with minibatch_size==1 into
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	228 single-example values in the result. Therefore the result of
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	229 iterating on the dataset itself gives a sequence of single examples
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	230 (whereas the result of iterating over minibatches gives in each
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	231 Example field an iterable object over the individual examples in
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	232 the minibatch).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	233 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	234 def __init__(self, minibatch_iterator):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	235 self.minibatch_iterator = minibatch_iterator
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	236 self.minibatch = None
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	237 def __iter__(self): #makes for loop work
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	238 return self
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	239 def next(self):
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	240 size1_minibatch = self.minibatch_iterator.next()
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	241 if not self.minibatch:
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	242 self.minibatch = Example(size1_minibatch.keys(),[value[0] for value in size1_minibatch.values()])
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	243 else:
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	244 self.minibatch._values = [value[0] for value in size1_minibatch.values()]
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	245 return self.minibatch
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	246
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	247 def next_index(self):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	248 return self.minibatch_iterator.next_index()
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	249
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	250 class MinibatchWrapAroundIterator(object):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	251 """
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	252 An iterator for minibatches that handles the case where we need to wrap around the
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	253 dataset because n_batches*minibatch_size > len(dataset). It is constructed from
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	254 a dataset that provides a minibatch iterator that does not need to handle that problem.
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	255 This class is a utility for dataset subclass writers, so that they do not have to handle
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	256 this issue multiple times, nor check that fieldnames are valid, nor handle the
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	257 empty fieldnames (meaning 'use all the fields').
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	258 """
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	259 def __init__(self,dataset,fieldnames,minibatch_size,n_batches,offset):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	260 self.dataset=dataset
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	261 self.fieldnames=fieldnames
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	262 self.minibatch_size=minibatch_size
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	263 self.n_batches=n_batches
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	264 self.n_batches_done=0
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	265 self.next_row=offset
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	266 self.L=len(dataset)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	267 self.offset=offset % self.L
98 7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	268 ds_nbatches = (self.L-self.next_row)/self.minibatch_size
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	269 if n_batches is not None:
98 7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	270 ds_nbatches = min(n_batches,ds_nbatches)
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	271 if fieldnames:
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	272 assert dataset.hasFields(*fieldnames)
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	273 else:
98 7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	274 self.fieldnames=dataset.fieldNames()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	275 self.iterator = self.dataset.minibatches_nowrap(self.fieldnames,self.minibatch_size, ds_nbatches,self.next_row)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	276
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	277 def __iter__(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	278 return self
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	279
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	280 def next_index(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	281 return self.next_row
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	282
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	283 def next(self):
43 e92244f30116 Corrected iterator logic errors bengioy@grenat.iro.umontreal.ca parents: 42 diff changeset	284 if self.n_batches and self.n_batches_done==self.n_batches:
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	285 raise StopIteration
101 a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	286 elif not self.n_batches and self.next_row ==self.L:
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	287 raise StopIteration
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	288 upper = self.next_row+self.minibatch_size
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	289 if upper <=self.L:
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	290 minibatch = self.iterator.next()
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	291 else:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	292 if not self.n_batches:
101 a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	293 upper=min(upper, self.L)
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	294 # if their is not a fixed number of batch, we continue to the end of the dataset.
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	295 # this can create a minibatch that is smaller then the minibatch_size
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	296 assert (self.L-self.next_row)<=self.minibatch_size
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	297 minibatch = self.dataset.minibatches_nowrap(self.fieldnames,self.L-self.next_row,1,self.next_row).next()
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	298 else:
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	299 # we must concatenate (vstack) the bottom and top parts of our minibatch
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	300 # first get the beginning of our minibatch (top of dataset)
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	301 first_part = self.dataset.minibatches_nowrap(self.fieldnames,self.L-self.next_row,1,self.next_row).next()
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	302 second_part = self.dataset.minibatches_nowrap(self.fieldnames,upper-self.L,1,0).next()
a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	303 minibatch = Example(self.fieldnames,
268 3f1cd8897fda reverting dataset James Bergstra <bergstrj@iro.umontreal.ca> parents: 266 diff changeset	304 [self.dataset.valuesVStack(name,[first_part[name],second_part[name]])
101 a1740a99b81f by default, in a minibatch without any fixed number of batchs, we need to finish at the end of the dataset. Now we return a minibatch at the end event if this minibacht size != the gived minibatch_size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 99 diff changeset	305 for name in self.fieldnames])
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	306 self.next_row=upper
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	307 self.n_batches_done+=1
43 e92244f30116 Corrected iterator logic errors bengioy@grenat.iro.umontreal.ca parents: 42 diff changeset	308 if upper >= self.L and self.n_batches:
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	309 self.next_row -= self.L
98 7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	310 ds_nbatches = (self.L-self.next_row)/self.minibatch_size
7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	311 if self.n_batches is not None:
7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	312 ds_nbatches = min(self.n_batches,ds_nbatches)
7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	313 self.iterator = self.dataset.minibatches_nowrap(self.fieldnames,self.minibatch_size,
7186e4f502d1 bugfix in DataSet.minibatch to correctly wrap in all corner case. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 95 diff changeset	314 ds_nbatches,self.next_row)
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	315 return DataSetFields(MinibatchDataSet(minibatch,self.dataset.valuesVStack,
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	316 self.dataset.valuesHStack),
74 b4159cbdc06b Fixed errors raised by test_dataset Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 73 diff changeset	317 minibatch.keys())
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	318
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	319
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	320 minibatches_fieldnames = None
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	321 minibatches_minibatch_size = 1
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	322 minibatches_n_batches = None
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	323 def minibatches(self,
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	324 fieldnames = minibatches_fieldnames,
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	325 minibatch_size = minibatches_minibatch_size,
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	326 n_batches = minibatches_n_batches,
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	327 offset = 0):
6 d5738b79089a Removed MinibatchIterator and instead made minibatch_size a field of all DataSets, bengioy@bengiomac.local parents: 5 diff changeset	328 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	329 Return an iterator that supports three forms of syntax:
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	330
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	331 for i in dataset.minibatches(None,**kwargs): ...
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	332
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	333 for i in dataset.minibatches([f1, f2, f3],**kwargs): ...
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	334
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	335 for i1, i2, i3 in dataset.minibatches([f1, f2, f3],**kwargs): ...
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	336
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	337 Using the first two syntaxes, "i" will be an indexable object, such as a list,
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	338 tuple, or Example instance. In both cases, i[k] is a list-like container
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	339 of a batch of current examples. In the second case, i[0] is
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	340 list-like container of the f1 field of a batch current examples, i[1] is
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	341 a list-like container of the f2 field, etc.
2 3fddb1c8f955 Rewrote DataSet interface and created FiniteDataSet interface. bengioy@bengiomac.local parents: 1 diff changeset	342
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	343 Using the first syntax, all the fields will be returned in "i".
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	344 Using the third syntax, i1, i2, i3 will be list-like containers of the
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	345 f1, f2, and f3 fields of a batch of examples on each loop iteration.
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	346
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	347 The minibatches iterator is expected to return upon each call to next()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	348 a DataSetFields object, which is a Example (indexed by the field names) whose
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	349 elements are iterable and indexable over the minibatch examples, and which keeps a pointer to
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	350 a sub-dataset that can be used to iterate over the individual examples
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	351 in the minibatch. Hence a minibatch can be converted back to a regular
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	352 dataset or its fields can be looked at individually (and possibly iterated over).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	353
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	354 PARAMETERS
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	355 - fieldnames (list of any type, default None):
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	356 The loop variables i1, i2, i3 (in the example above) should contain the
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	357 f1, f2, and f3 fields of the current batch of examples. If None, the
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	358 derived class can choose a default, e.g. all fields.
16 813723310d75 commenting bergstrj@iro.umontreal.ca parents: 15 11 diff changeset	359
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	360 - minibatch_size (integer, default 1)
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	361 On every iteration, the variables i1, i2, i3 will have
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	362 exactly minibatch_size elements. e.g. len(i1) == minibatch_size
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	363
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	364 - n_batches (integer, default None)
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	365 The iterator will loop exactly this many times, and then stop. If None,
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	366 the derived class can choose a default. If (-1), then the returned
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	367 iterator should support looping indefinitely.
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	368
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	369 - offset (integer, default 0)
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	370 The iterator will start at example 'offset' in the dataset, rather than the default.
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	371
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	372 Note: A list-like container is something like a tuple, list, numpy.ndarray or
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	373 any other object that supports integer indexing and slicing.
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	374
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	375 """
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	376 return DataSet.MinibatchWrapAroundIterator(self,fieldnames,minibatch_size,n_batches,offset)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	377
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	378 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	379 """
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	380 This is the minibatches iterator generator that sub-classes must define.
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	381 It does not need to worry about wrapping around multiple times across the dataset,
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	382 as this is handled by MinibatchWrapAroundIterator when DataSet.minibatches() is called.
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	383 The next() method of the returned iterator does not even need to worry about
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	384 the termination condition (as StopIteration will be raised by DataSet.minibatches
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	385 before an improper call to minibatches_nowrap's next() is made).
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	386 That next() method can assert that its next row will always be within [0,len(dataset)).
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	387 The iterator returned by minibatches_nowrap does not need to implement
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	388 a next_index() method either, as this will be provided by MinibatchWrapAroundIterator.
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	389 """
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	390 raise AbstractFunction()
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	391
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	392 def is_unbounded(self):
b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	393 """
b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	394 Tests whether a dataset is unbounded (e.g. a stream).
b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	395 """
123 8b520423d4ee bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 110 diff changeset	396 return len(self)==maxint
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	397
26 672fe4b23032 Fixed dataset errors so that _test_dataset.py works again. bengioy@grenat.iro.umontreal.ca parents: 23 diff changeset	398 def hasFields(self,*fieldnames):
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	399 """
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	400 Return true if the given field name (or field names, if multiple arguments are
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	401 given) is recognized by the DataSet (i.e. can be used as a field name in one
b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	402 of the iterators).
29 46c5c90019c2 Changed apply_function so that it propagates methods of the source. bengioy@grenat.iro.umontreal.ca parents: 28 diff changeset	403
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	404 The default implementation may be inefficient (O(# fields in dataset)), as it calls the fieldNames()
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	405 method. Many datasets may store their field names in a dictionary, which would allow more efficiency.
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	406 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	407 return len(unique_elements_list_intersection(fieldnames,self.fieldNames()))>0
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	408
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	409 def fieldNames(self):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	410 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	411 Return the list of field names that are supported by the iterators,
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	412 and for which hasFields(fieldname) would return True.
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	413 """
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	414 raise AbstractFunction()
759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	415
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	416 def __call__(self,*fieldnames):
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	417 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	418 Return a dataset that sees only the fields whose name are specified.
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	419 """
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	420 assert self.hasFields(*fieldnames)
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	421 return self.fields(*fieldnames).examples()
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	422
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	423 def fields(self,*fieldnames):
29 46c5c90019c2 Changed apply_function so that it propagates methods of the source. bengioy@grenat.iro.umontreal.ca parents: 28 diff changeset	424 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	425 Return a DataSetFields object associated with this dataset.
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	426 """
74 b4159cbdc06b Fixed errors raised by test_dataset Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 73 diff changeset	427 return DataSetFields(self,fieldnames)
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	428
269 fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	429 def getitem_key(self, fieldname):
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	430 """A not-so-well thought-out place to put code that used to be in
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	431 getitem.
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	432 """
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	433 #removing as per discussion June 4. --JSB
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	434
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	435 i = fieldname
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	436 # else check for a fieldname
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	437 if self.hasFields(i):
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	438 return self.minibatches(fieldnames=[i],minibatch_size=len(self),n_batches=1,offset=0).next()[0]
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	439 # else we are trying to access a property of the dataset
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	440 assert i in self.__dict__ # else it means we are trying to access a non-existing property
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	441 return self.__dict__[i]
fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	442
2 3fddb1c8f955 Rewrote DataSet interface and created FiniteDataSet interface. bengioy@bengiomac.local parents: 1 diff changeset	443 def __getitem__(self,i):
28 541a273bc89f Removed __array__ method from dataset, whose bengioy@grenat.iro.umontreal.ca parents: 26 diff changeset	444 """
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	445 @rtype: Example
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	446 @returns: single or multiple examples
1 2cd82666b9a7 Added statscollector and started writing dataset and learner. bengioy@esprit.iro.umontreal.ca parents: 0 diff changeset	447
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	448 @type i: integer or slice or <iterable> of integers
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	449 @param i:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	450 dataset[i] returns the (i+1)-th example of the dataset.
309 923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	451 dataset[i:j] returns a LookupList with examples i,i+1,...,j-1.
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	452 dataset[i:j:s] returns a LookupList with examples i,i+2,i+4...,j-2.
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	453 dataset[[i1,i2,..,in]] returns a LookupList with examples i1,i2,...,in.
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	454
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	455 @note:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	456 Some stream datasets may be unable to implement random access, i.e.
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	457 arbitrary slicing/indexing because they can only iterate through
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	458 examples one or a minibatch at a time and do not actually store or keep
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	459 past (or future) examples.
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	460
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	461 The default implementation of getitem uses the minibatches iterator
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	462 to obtain one example, one slice, or a list of examples. It may not
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	463 always be the most efficient way to obtain the result, especially if
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	464 the data are actually stored in a memory array.
28 541a273bc89f Removed __array__ method from dataset, whose bengioy@grenat.iro.umontreal.ca parents: 26 diff changeset	465 """
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	466
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	467 if type(i) is int:
309 923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	468 assert i >= 0 # TBM: see if someone complains and want negative i
313 009ce84e9f52 behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 310 diff changeset	469 if i >= len(self) :
009ce84e9f52 behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 310 diff changeset	470 raise IndexError
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	471 i_batch = self.minibatches_nowrap(self.fieldNames(),
293 4bfdda107a17 still merging James Bergstra <bergstrj@iro.umontreal.ca> parents: 292 diff changeset	472 minibatch_size=1, n_batches=1, offset=i)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	473 return DataSet.MinibatchToSingleExampleIterator(i_batch).next()
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	474
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	475 #if i is a contiguous slice
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	476 if type(i) is slice and (i.step in (None, 1)):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	477 offset = 0 if i.start is None else i.start
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	478 upper_bound = len(self) if i.stop is None else i.stop
313 009ce84e9f52 behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 310 diff changeset	479 upper_bound = min(len(self) , upper_bound)
309 923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	480 #return MinibatchDataSet(self.minibatches_nowrap(self.fieldNames(),
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	481 # minibatch_size=upper_bound - offset,
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	482 # n_batches=1,
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	483 # offset=offset).next())
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	484 # now returns a LookupList
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	485 return self.minibatches_nowrap(self.fieldNames(),
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	486 minibatch_size=upper_bound - offset,
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	487 n_batches=1,
309 923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	488 offset=offset).next()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	489
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	490 # if slice has a step param, convert it to list and handle it with the
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	491 # list code
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	492 if type(i) is slice:
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	493 offset = 0 if i.start is None else i.start
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	494 upper_bound = len(self) if i.stop is None else i.stop
314 105b54ac8260 bug fixed concerning the slicing, now ds[0:len(ds) + 1000 : 2] is accepted, same a python list Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 313 diff changeset	495 upper_bound = min(len(self) , upper_bound)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	496 i = list(range(offset, upper_bound, i.step))
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	497
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	498 # handle tuples, arrays, lists
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	499 if hasattr(i, '__getitem__'):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	500 for idx in i:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	501 #dis-allow nested slices
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	502 if not isinstance(idx, int):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	503 raise TypeError(idx)
313 009ce84e9f52 behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 310 diff changeset	504 if idx >= len(self) :
009ce84e9f52 behaviour is now the same as a list in pylearn, so if len(ds) = 10, ds[10] raise an IndexError, same thing for ds[[1,10]], and ds[0:14:1] returns 10 elements Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 310 diff changeset	505 raise IndexError
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	506 # call back into self.__getitem__
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	507 examples = [self.minibatches_nowrap(self.fieldNames(),
293 4bfdda107a17 still merging James Bergstra <bergstrj@iro.umontreal.ca> parents: 292 diff changeset	508 minibatch_size=1, n_batches=1, offset=ii).next()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	509 for ii in i]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	510 # re-index the fields in each example by field instead of by example
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	511 field_values = [[] for blah in self.fieldNames()]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	512 for e in examples:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	513 for f,v in zip(field_values, e):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	514 f.append(v)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	515 #build them into a LookupList (a.ka. Example)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	516 zz = zip(self.fieldNames(),field_values)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	517 vst = [self.valuesVStack(fieldname,field_values) for fieldname,field_values in zz]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	518 example = Example(self.fieldNames(), vst)
309 923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	519 #return MinibatchDataSet(example, self.valuesVStack, self.valuesHStack)
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	520 # now returns a LookupList
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	521 return example
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	522
923de30457f0 get item now returns LookupLists Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 296 diff changeset	523 # what in the world is i?
269 fdce496c3b56 deprecating __getitem__[fieldname] syntax James Bergstra <bergstrj@iro.umontreal.ca> parents: 268 diff changeset	524 raise TypeError(i, type(i))
22 b6b36f65664f Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet, bengioy@esprit.iro.umontreal.ca parents: 20 diff changeset	525
310 ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	526
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	527 """
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	528 Enables the call dataset.subset[a:b:c] that will return a DataSet
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	529 around the examples returned by __getitem__(slice(a,b,c))
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	530
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	531 @SEE DataSet.__getsubset(self)
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	532 """
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	533 subset = property(lambda s : s.__getsubset(),doc="returns a subset as a DataSet")
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	534
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	535
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	536 def __getsubset(self) :
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	537 """
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	538 Enables the call data.subset[a:b:c], returns a DataSet.
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	539 Default implementation is a simple wrap around __getitem__() using MinibatchDataSet.
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	540
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	541 @RETURN DataSet
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	542 @SEE DataSet.subset = property(lambda s : s.__getsubset())
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	543 """
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	544 _self = self
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	545 class GetSliceReturnsDataSet(object) :
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	546 def __getitem__(self,slice) :
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	547 return MinibatchDataSet(_self.__getitem__(slice))
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	548 return GetSliceReturnsDataSet()
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	549
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	550
ebccfd05ccd5 dataset.subset implemented Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 309 diff changeset	551
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	552 def valuesHStack(self,fieldnames,fieldvalues):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	553 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	554 Return a value that corresponds to concatenating (horizontally) several field values.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	555 This can be useful to merge some fields. The implementation of this operation is likely
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	556 to involve a copy of the original values. When the values are numpy arrays, the
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	557 result should be numpy.hstack(values). If it makes sense, this operation should
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	558 work as well when each value corresponds to multiple examples in a minibatch
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	559 e.g. if each value is a Ni-vector and a minibatch of length L is a LxNi matrix,
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	560 then the result should be a Lx(N1+N2+..) matrix equal to numpy.hstack(values).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	561 The default is to use numpy.hstack for numpy.ndarray values, and a list
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	562 pointing to the original values for other data types.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	563 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	564 all_numpy=True
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	565 for value in fieldvalues:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	566 if not type(value) is numpy.ndarray:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	567 all_numpy=False
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	568 if all_numpy:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	569 return numpy.hstack(fieldvalues)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	570 # the default implementation of horizontal stacking is to put values in a list
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	571 return fieldvalues
26 672fe4b23032 Fixed dataset errors so that _test_dataset.py works again. bengioy@grenat.iro.umontreal.ca parents: 23 diff changeset	572
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	573 def valuesVStack(self,fieldname,values):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	574 """
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	575 @param fieldname: the name of the field from which the values were taken
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	576 @type fieldname: any type
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	577
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	578 @param values: bits near the beginning or end of the dataset
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	579 @type values: list of minibatches (returned by minibatch_nowrap)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	580
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	581 @return: the concatenation (stacking) of the values
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	582 @rtype: something suitable as a minibatch field
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	583 """
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	584 rval = []
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	585 for v in values:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	586 rval.extend(v)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	587 return rval
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	588
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	589 def __or__(self,other):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	590 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	591 dataset1 \| dataset2 returns a dataset whose list of fields is the concatenation of the list of
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	592 fields of the argument datasets. This only works if they all have the same length.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	593 """
135 0d8e721cc63c Fixed bugs in dataset to make test_mlp.py work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 134 diff changeset	594 return HStackedDataSet([self,other])
3 378b68d5c4ad Added first (untested) version of ArrayDataSet bengioy@bengiomac.local parents: 2 diff changeset	595
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	596 def __and__(self,other):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	597 """
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	598 dataset1 & dataset2 is a dataset that concatenates the examples from the argument datasets
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	599 (and whose length is the sum of the length of the argument datasets). This only
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	600 works if they all have the same fields.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	601 """
135 0d8e721cc63c Fixed bugs in dataset to make test_mlp.py work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 134 diff changeset	602 return VStackedDataSet([self,other])
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	603
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	604 def hstack(datasets):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	605 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	606 hstack(dataset1,dataset2,...) returns dataset1 \| datataset2 \| ...
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	607 which is a dataset whose fields list is the concatenation of the fields
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	608 of the individual datasets.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	609 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	610 assert len(datasets)>0
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	611 if len(datasets)==1:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	612 return datasets[0]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	613 return HStackedDataSet(datasets)
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	614
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	615 def vstack(datasets):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	616 """
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	617 vstack(dataset1,dataset2,...) returns dataset1 & datataset2 & ...
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	618 which is a dataset which iterates first over the examples of dataset1, then
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	619 over those of dataset2, etc.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	620 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	621 assert len(datasets)>0
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	622 if len(datasets)==1:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	623 return datasets[0]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	624 return VStackedDataSet(datasets)
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	625
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	626 class FieldsSubsetDataSet(DataSet):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	627 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	628 A sub-class of L{DataSet} that selects a subset of the fields.
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	629 """
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	630 def __init__(self,src,fieldnames):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	631 self.src=src
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	632 self.fieldnames=fieldnames
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	633 assert src.hasFields(*fieldnames)
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	634 self.valuesHStack = src.valuesHStack
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	635 self.valuesVStack = src.valuesVStack
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	636
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	637 def __len__(self): return len(self.src)
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	638
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	639 def fieldNames(self):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	640 return self.fieldnames
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	641
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	642 def __iter__(self):
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	643 class FieldsSubsetIterator(object):
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	644 def __init__(self,ds):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	645 self.ds=ds
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	646 self.src_iter=ds.src.__iter__()
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	647 self.example=None
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	648 def __iter__(self): return self
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	649 def next(self):
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	650 complete_example = self.src_iter.next()
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	651 if self.example:
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	652 self.example._values=[complete_example[field]
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	653 for field in self.ds.fieldnames]
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	654 else:
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	655 self.example=Example(self.ds.fieldnames,
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	656 [complete_example[field] for field in self.ds.fieldnames])
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	657 return self.example
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	658 return FieldsSubsetIterator(self)
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	659
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	660 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	661 assert self.hasFields(*fieldnames)
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	662 return self.src.minibatches_nowrap(fieldnames,minibatch_size,n_batches,offset)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	663 def dontuse__getitem__(self,i):
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	664 return FieldsSubsetDataSet(self.src[i],self.fieldnames)
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	665
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	666
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	667 class DataSetFields(Example):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	668 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	669 Although a L{DataSet} iterates over examples (like rows of a matrix), an associated
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	670 DataSetFields iterates over fields (like columns of a matrix), and can be understood
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	671 as a transpose of the associated dataset.
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	672
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	673 To iterate over fields, one can do
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	674 * for fields in dataset.fields()
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	675 * for fields in dataset(field1,field2,...).fields() to select a subset of fields
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	676 * for fields in dataset.fields(field1,field2,...) to select a subset of fields
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	677 and each of these fields is iterable over the examples:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	678 * for field_examples in dataset.fields():
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	679 for example_value in field_examples:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	680 ...
241 ddb88a8e9fd2 If I understand properly, the length of an unbounded stream is sys.maxint delallea@opale.iro.umontreal.ca parents: 231 diff changeset	681 but when the dataset is a stream (unbounded length), it is not recommended to do
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	682 such things because the underlying dataset may refuse to access the different fields in
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	683 an unsynchronized ways. Hence the fields() method is illegal for streams, by default.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	684 The result of fields() is a DataSetFields object, which iterates over fields,
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	685 and whose elements are iterable over examples. A DataSetFields object can
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	686 be turned back into a DataSet with its examples() method:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	687 dataset2 = dataset1.fields().examples()
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	688 and dataset2 should behave exactly like dataset1 (in fact by default dataset2==dataset1).
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	689
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	690 DataSetFields can be concatenated vertically or horizontally. To be consistent with
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	691 the syntax used for DataSets, the \| concatenates the fields and the & concatenates
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	692 the examples.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	693 """
74 b4159cbdc06b Fixed errors raised by test_dataset Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 73 diff changeset	694 def __init__(self,dataset,fieldnames):
65 d48eba49a2f4 fixed the infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 64 diff changeset	695 original_dataset=dataset
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	696 if not fieldnames:
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	697 fieldnames=dataset.fieldNames()
274 ed70580f2324 bugfix in FieldSubsetDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 273 diff changeset	698 elif not list(fieldnames)==list(dataset.fieldNames()):
ed70580f2324 bugfix in FieldSubsetDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 273 diff changeset	699 #we must cast to list, othersize('x','y')!=['x','y']
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	700 dataset = FieldsSubsetDataSet(dataset,fieldnames)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	701 assert dataset.hasFields(*fieldnames)
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	702 self.dataset=dataset
66 dde1fb1b63ba fixed test and removed print Frederic Bastien <bastienf@iro.umontreal.ca> parents: 65 diff changeset	703
64 863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	704 if isinstance(dataset,MinibatchDataSet):
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	705 Example.__init__(self,fieldnames,list(dataset._fields))
65 d48eba49a2f4 fixed the infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 64 diff changeset	706 elif isinstance(original_dataset,MinibatchDataSet):
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	707 Example.__init__(self,fieldnames,
65 d48eba49a2f4 fixed the infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 64 diff changeset	708 [original_dataset._fields[field]
d48eba49a2f4 fixed the infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 64 diff changeset	709 for field in fieldnames])
64 863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	710 else:
863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	711 minibatch_iterator = dataset.minibatches(fieldnames,
863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	712 minibatch_size=len(dataset),
863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	713 n_batches=1)
863da25a60f1 trying to fix infinite loop Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 62 diff changeset	714 minibatch=minibatch_iterator.next()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	715 Example.__init__(self,fieldnames,minibatch)
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	716
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	717 def examples(self):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	718 return self.dataset
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	719
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	720 def __or__(self,other):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	721 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	722 fields1 \| fields2 is a DataSetFields that whose list of examples is the concatenation
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	723 of the list of examples of DataSetFields fields1 and fields2.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	724 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	725 return (self.examples() + other.examples()).fields()
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	726
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	727 def __and__(self,other):
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	728 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	729 fields1 + fields2 is a DataSetFields that whose list of fields is the concatenation
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	730 of the fields of DataSetFields fields1 and fields2.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	731 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	732 return (self.examples() \| other.examples()).fields()
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	733
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	734
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	735 class MinibatchDataSet(DataSet):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	736 """
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	737 Turn a L{Example} of same-length (iterable) fields into an example-iterable dataset.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	738 Each element of the lookup-list should be an iterable and sliceable, all of the same length.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	739 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	740 def __init__(self,fields_lookuplist,values_vstack=DataSet().valuesVStack,
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	741 values_hstack=DataSet().valuesHStack):
17 759d17112b23 more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests bergstrj@iro.umontreal.ca parents: 16 12 diff changeset	742 """
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	743 The user can (and generally should) also provide values_vstack(fieldname,fieldvalues)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	744 and a values_hstack(fieldnames,fieldvalues) functions behaving with the same
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	745 semantics as the DataSet methods of the same name (but without the self argument).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	746 """
211 bd728c83faff in __get__, problem if the i.stop was None, i being the slice, added one line replacing None by the len(self) Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 203 diff changeset	747
61 a8b70a9117ad bugfix: in MinibatchDataSet renamed the class variable fields to _fields as parent class have a function called field. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 60 diff changeset	748 self._fields=fields_lookuplist
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	749 assert len(fields_lookuplist)>0
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	750 self.length=len(fields_lookuplist[0])
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	751 for field in fields_lookuplist[1:]:
223 517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	752 if self.length != len(field) :
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	753 print 'self.length = ',self.length
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	754 print 'len(field) = ', len(field)
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	755 print 'self._fields.keys() = ', self._fields.keys()
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	756 print 'field=',field
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	757 print 'fields_lookuplist=', fields_lookuplist
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	758 assert self.length==len(field)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	759 self.valuesVStack=values_vstack
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	760 self.valuesHStack=values_hstack
3 378b68d5c4ad Added first (untested) version of ArrayDataSet bengioy@bengiomac.local parents: 2 diff changeset	761
378b68d5c4ad Added first (untested) version of ArrayDataSet bengioy@bengiomac.local parents: 2 diff changeset	762 def __len__(self):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	763 return self.length
28 541a273bc89f Removed __array__ method from dataset, whose bengioy@grenat.iro.umontreal.ca parents: 26 diff changeset	764
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	765 def dontuse__getitem__(self,i):
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	766 if type(i) in (slice,list):
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	767 return DataSetFields(MinibatchDataSet(
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	768 Example(self._fields.keys(),[field[i] for field in self._fields])),self.fieldNames())
40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	769 if type(i) is int:
85 7b1f3ad3d60b bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 83 diff changeset	770 return Example(self._fields.keys(),[field[i] for field in self._fields])
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	771 if self.hasFields(i):
61 a8b70a9117ad bugfix: in MinibatchDataSet renamed the class variable fields to _fields as parent class have a function called field. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 60 diff changeset	772 return self._fields[i]
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	773 assert i in self.__dict__ # else it means we are trying to access a non-existing property
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	774 return self.__dict__[i]
11 be128b9127c8 Debugged (to the extent of my tests) the new version of dataset bengioy@esprit.iro.umontreal.ca parents: 9 diff changeset	775
29 46c5c90019c2 Changed apply_function so that it propagates methods of the source. bengioy@grenat.iro.umontreal.ca parents: 28 diff changeset	776 def fieldNames(self):
61 a8b70a9117ad bugfix: in MinibatchDataSet renamed the class variable fields to _fields as parent class have a function called field. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 60 diff changeset	777 return self._fields.keys()
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	778
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	779 def hasFields(self,*fieldnames):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	780 for fieldname in fieldnames:
61 a8b70a9117ad bugfix: in MinibatchDataSet renamed the class variable fields to _fields as parent class have a function called field. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 60 diff changeset	781 if fieldname not in self._fields.keys():
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	782 return False
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	783 return True
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	784
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	785 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
223 517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	786 #@TODO bug somewhere here, fieldnames doesnt seem to be well handled
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	787 class Iterator(object):
223 517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	788 def __init__(self,ds,fieldnames):
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	789 # tbm: added two next lines to handle fieldnames
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	790 if fieldnames is None: fieldnames = ds._fields.keys()
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	791 self.fieldnames = fieldnames
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	792
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	793 self.ds=ds
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	794 self.next_example=offset
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	795 assert minibatch_size >= 0
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	796 if offset+minibatch_size > ds.length:
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	797 raise NotImplementedError()
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	798 def __iter__(self):
266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	799 return self
266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	800 def next(self):
61 a8b70a9117ad bugfix: in MinibatchDataSet renamed the class variable fields to _fields as parent class have a function called field. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 60 diff changeset	801 upper = self.next_example+minibatch_size
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	802 assert upper<=self.ds.length
223 517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	803 #minibatch = Example(self.ds._fields.keys(),
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	804 # [field[self.next_example:upper]
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	805 # for field in self.ds._fields])
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	806 # tbm: modif to use fieldnames
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	807 values = []
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	808 for f in self.fieldnames :
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	809 #print 'we have field',f,'in fieldnames'
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	810 values.append( self.ds._fields[f][self.next_example:upper] )
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	811 minibatch = Example(self.fieldnames,values)
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	812 #print minibatch
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	813 self.next_example+=minibatch_size
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	814 return minibatch
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	815
223 517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	816 # tbm: added fieldnames to handle subset of fieldnames
517364d48ae0 should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 218 diff changeset	817 return Iterator(self,fieldnames)
20 266c68cb6136 Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works. bengioy@bengiomac.local parents: 19 diff changeset	818
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	819 class HStackedDataSet(DataSet):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	820 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	821 A L{DataSet} that wraps several datasets and shows a view that includes all their fields,
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	822 i.e. whose list of fields is the concatenation of their lists of fields.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	823
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	824 If a field name is found in more than one of the datasets, then either an error is
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	825 raised or the fields are renamed (either by prefixing the __name__ attribute
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	826 of the dataset + ".", if it exists, or by suffixing the dataset index in the argument list).
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	827
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	828 @todo: automatically detect a chain of stacked datasets due to A \| B \| C \| D ...
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	829 """
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	830 def __init__(self,datasets,accept_nonunique_names=False,description=None,field_types=None):
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	831 DataSet.__init__(self,description,field_types)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	832 self.datasets=datasets
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	833 self.accept_nonunique_names=accept_nonunique_names
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	834 self.fieldname2dataset={}
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	835
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	836 def rename_field(fieldname,dataset,i):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	837 if hasattr(dataset,"__name__"):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	838 return dataset.__name__ + "." + fieldname
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	839 return fieldname+"."+str(i)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	840
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	841 # make sure all datasets have the same length and unique field names
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	842 self.length=None
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	843 names_to_change=[]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	844 for i in xrange(len(datasets)):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	845 dataset = datasets[i]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	846 length=len(dataset)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	847 if self.length:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	848 assert self.length==length
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	849 else:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	850 self.length=length
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	851 for fieldname in dataset.fieldNames():
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	852 if fieldname in self.fieldname2dataset: # name conflict!
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	853 if accept_nonunique_names:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	854 fieldname=rename_field(fieldname,dataset,i)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	855 names2change.append((fieldname,i))
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	856 else:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	857 raise ValueError("Incompatible datasets: non-unique field name = "+fieldname)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	858 self.fieldname2dataset[fieldname]=i
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	859 for fieldname,i in names_to_change:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	860 del self.fieldname2dataset[fieldname]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	861 self.fieldname2dataset[rename_field(fieldname,self.datasets[i],i)]=i
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	862
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	863 def hasFields(self,*fieldnames):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	864 for fieldname in fieldnames:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	865 if not fieldname in self.fieldname2dataset:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	866 return False
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	867 return True
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	868
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	869 def fieldNames(self):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	870 return self.fieldname2dataset.keys()
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	871
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	872 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	873
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	874 class HStackedIterator(object):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	875 def __init__(self,hsds,iterators):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	876 self.hsds=hsds
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	877 self.iterators=iterators
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	878 def __iter__(self):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	879 return self
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	880 def next(self):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	881 # concatenate all the fields of the minibatches
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	882 l=Example()
140 488ad6398b12 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 136 diff changeset	883 for iter in self.iterators:
488ad6398b12 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 136 diff changeset	884 l.append_lookuplist(iter.next())
488ad6398b12 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 136 diff changeset	885 return l
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	886
125 7d8b3d6dd4e9 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 123 diff changeset	887 assert self.hasFields(*fieldnames)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	888 # find out which underlying datasets are necessary to service the required fields
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	889 # and construct corresponding minibatch iterators
140 488ad6398b12 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 136 diff changeset	890 if fieldnames and fieldnames!=self.fieldNames():
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	891 datasets=set([])
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	892 fields_in_dataset=dict([(dataset,[]) for dataset in datasets])
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	893 for fieldname in fieldnames:
136 fa752c55aa09 typo Frederic Bastien <bastienf@iro.umontreal.ca> parents: 125 diff changeset	894 dataset=self.datasets[self.fieldname2dataset[fieldname]]
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	895 datasets.add(dataset)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	896 fields_in_dataset[dataset].append(fieldname)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	897 datasets=list(datasets)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	898 iterators=[dataset.minibatches(fields_in_dataset[dataset],minibatch_size,n_batches,offset)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	899 for dataset in datasets]
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	900 else:
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	901 datasets=self.datasets
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	902 iterators=[dataset.minibatches(None,minibatch_size,n_batches,offset) for dataset in datasets]
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	903 return HStackedIterator(self,iterators)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	904
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	905
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	906 def untested_valuesVStack(self,fieldname,fieldvalues):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	907 return self.datasets[self.fieldname2dataset[fieldname]].valuesVStack(fieldname,fieldvalues)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	908
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	909 def untested_valuesHStack(self,fieldnames,fieldvalues):
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	910 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	911 We will use the sub-dataset associated with the first fieldname in the fieldnames list
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	912 to do the work, hoping that it can cope with the other values (i.e. won't care
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	913 about the incompatible fieldnames). Hence this heuristic will always work if
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	914 all the fieldnames are of the same sub-dataset.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	915 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	916 return self.datasets[self.fieldname2dataset[fieldnames[0]]].valuesHStack(fieldnames,fieldvalues)
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	917
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	918 class VStackedDataSet(DataSet):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	919 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	920 A L{DataSet} that wraps several datasets and shows a view that includes all their examples,
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	921 in the order provided. This clearly assumes that they all have the same field names
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	922 and all (except possibly the last one) are of finite length.
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	923
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	924 @todo: automatically detect a chain of stacked datasets due to A + B + C + D ...
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	925 """
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	926 def __init__(self,datasets):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	927 self.datasets=datasets
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	928 self.length=0
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	929 self.index2dataset={}
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	930 assert len(datasets)>0
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	931 fieldnames = datasets[-1].fieldNames()
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	932 self.datasets_start_row=[]
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	933 # We use this map from row index to dataset index for constant-time random access of examples,
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	934 # to avoid having to search for the appropriate dataset each time and slice is asked for.
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	935 for dataset,k in enumerate(datasets[0:-1]):
48 b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	936 assert dataset.is_unbounded() # All VStacked datasets (except possibly the last) must be bounded (have a length).
b6730f9a336d Fixing MinibatchDataSet getitem bengioy@grenat.iro.umontreal.ca parents: 46 diff changeset	937 L=len(dataset)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	938 for i in xrange(L):
438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	939 self.index2dataset[self.length+i]=k
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	940 self.datasets_start_row.append(self.length)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	941 self.length+=L
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	942 assert dataset.fieldNames()==fieldnames
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	943 self.datasets_start_row.append(self.length)
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	944 self.length+=len(datasets[-1])
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	945 # If length is very large, we should use a more memory-efficient mechanism
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	946 # that does not store all indices
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	947 if self.length>1000000:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	948 # 1 million entries would require about 60 meg for the index2dataset map
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	949 # TODO
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	950 print "A more efficient mechanism for index2dataset should be implemented"
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	951
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	952 def __len__(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	953 return self.length
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	954
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	955 def fieldNames(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	956 return self.datasets[0].fieldNames()
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	957
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	958 def hasFields(self,*fieldnames):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	959 return self.datasets[0].hasFields(*fieldnames)
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	960
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	961 def locate_row(self,row):
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	962 """Return (dataset_index, row_within_dataset) for global row number"""
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	963 dataset_index = self.index2dataset[row]
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	964 row_within_dataset = self.datasets_start_row[dataset_index]
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	965 return dataset_index, row_within_dataset
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	966
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	967 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	968
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	969 class VStackedIterator(object):
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	970 def __init__(self,vsds):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	971 self.vsds=vsds
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	972 self.next_row=offset
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	973 self.next_dataset_index,self.next_dataset_row=self.vsds.locate_row(offset)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	974 self.current_iterator,self.n_left_at_the_end_of_ds,self.n_left_in_mb= \
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	975 self.next_iterator(vsds.datasets[0],offset,n_batches)
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	976
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	977 def next_iterator(self,dataset,starting_offset,batches_left):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	978 L=len(dataset)
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	979 ds_nbatches = (L-starting_offset)/minibatch_size
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	980 if batches_left is not None:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	981 ds_nbatches = max(batches_left,ds_nbatches)
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	982 if minibatch_size>L:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	983 ds_minibatch_size=L
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	984 n_left_in_mb=minibatch_size-L
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	985 ds_nbatches=1
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	986 else:
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	987 n_left_in_mb=0
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	988 return dataset.minibatches(fieldnames,minibatch_size,ds_nbatches,starting_offset), \
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	989 L-(starting_offset+ds_nbatches*minibatch_size), n_left_in_mb
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	990
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	991 def move_to_next_dataset(self):
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	992 if self.n_left_at_the_end_of_ds>0:
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	993 self.current_iterator,self.n_left_at_the_end_of_ds,self.n_left_in_mb= \
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	994 self.next_iterator(vsds.datasets[self.next_dataset_index],
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	995 self.n_left_at_the_end_of_ds,1)
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	996 else:
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	997 self.next_dataset_index +=1
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	998 if self.next_dataset_index==len(self.vsds.datasets):
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	999 self.next_dataset_index = 0
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	1000 self.current_iterator,self.n_left_at_the_end_of_ds,self.n_left_in_mb= \
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	1001 self.next_iterator(vsds.datasets[self.next_dataset_index],starting_offset,n_batches)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1002
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1003 def __iter__(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1004 return self
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1005
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1006 def next(self):
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1007 dataset=self.vsds.datasets[self.next_dataset_index]
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1008 mb = self.next_iterator.next()
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1009 if self.n_left_in_mb:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1010 extra_mb = []
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1011 while self.n_left_in_mb>0:
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1012 self.move_to_next_dataset()
73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1013 extra_mb.append(self.next_iterator.next())
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1014 mb = Example(fieldnames,
40 88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	1015 [dataset.valuesVStack(name,
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	1016 [mb[name]]+[b[name] for b in extra_mb])
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	1017 for name in fieldnames])
88fd1cce08b9 replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets bengioy@esprit.iro.umontreal.ca parents: 39 diff changeset	1018
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1019 self.next_row+=minibatch_size
38 d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	1020 self.next_dataset_row+=minibatch_size
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	1021 if self.next_row+minibatch_size>len(dataset):
d637ad8f7352 Finished first untested version of VStackedDataset bengioy@esprit.iro.umontreal.ca parents: 37 diff changeset	1022 self.move_to_next_dataset()
44 5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	1023 return examples
5a85fda9b19b Fixed some more iterator bugs bengioy@grenat.iro.umontreal.ca parents: 43 diff changeset	1024 return VStackedIterator(self)
37 73c4212ba5b3 Factored the minibatch-writing code into an iterator class inside DataSet bengioy@esprit.iro.umontreal.ca parents: 36 diff changeset	1025
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1026 class ArrayFieldsDataSet(DataSet):
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1027 """
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1028 Virtual super-class of datasets whose field values are numpy array,
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1029 thus defining valuesHStack and valuesVStack for sub-classes.
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1030 """
268 3f1cd8897fda reverting dataset James Bergstra <bergstrj@iro.umontreal.ca> parents: 266 diff changeset	1031 def __init__(self,description=None,field_types=None):
3f1cd8897fda reverting dataset James Bergstra <bergstrj@iro.umontreal.ca> parents: 266 diff changeset	1032 DataSet.__init__(self,description,field_types)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1033 def untested_valuesHStack(self,fieldnames,fieldvalues):
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1034 """Concatenate field values horizontally, e.g. two vectors
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1035 become a longer vector, two matrices become a wider matrix, etc."""
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1036 return numpy.hstack(fieldvalues)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1037 def untested_valuesVStack(self,fieldname,values):
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1038 """Concatenate field values vertically, e.g. two vectors
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1039 become a two-row matrix, two matrices become a longer matrix, etc."""
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1040 return numpy.vstack(values)
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1041
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1042
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1043
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1044 class NArraysDataSet(ArrayFieldsDataSet) :
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1045 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1046 An NArraysDataSet stores fields that are numpy tensor, whose first axis
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1047 iterates over examples. It's a generalization of ArrayDataSet.
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1048 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1049 #@TODO not completely implemented yet
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1050 def __init__(self, data_arrays, fieldnames, **kwargs) :
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1051 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1052 Construct an NArraysDataSet from a list of numpy tensor (data_arrays) and a list
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1053 of fieldnames. The number of arrays must be the same as the number of
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1054 fieldnames. Each set of numpy tensor must have the same first dimension (first
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1055 axis) corresponding to the number of examples.
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1056
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1057 Every tensor is treated as a numpy matrix (using numpy.asmatrix)
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1058 """
318 e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1059 ArrayFieldsDataSet.__init__(self,**kwargs)
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1060 assert len(data_arrays) == len(fieldnames)
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1061 assert len(fieldnames) > 0
319 6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1062 all_matrix_sizes = map(lambda x : numpy.asmatrix(x).shape[0] , data_arrays)
6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1063 num_examples = max(all_matrix_sizes)
320 b1da46b9b901 bug fixed when initializing NArraysDataSet with 1-d arrays. This now raises an Exception Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 319 diff changeset	1064 if num_examples == 1 :
b1da46b9b901 bug fixed when initializing NArraysDataSet with 1-d arrays. This now raises an Exception Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 319 diff changeset	1065 # problem, do we transpose all arrays? is there only one example?
b1da46b9b901 bug fixed when initializing NArraysDataSet with 1-d arrays. This now raises an Exception Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 319 diff changeset	1066 raise Exception("wrong initialization, unknow behaviour with 1-d arrays")
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1067 self._fieldnames = fieldnames
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1068 self._datas = []
318 e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1069 for k in range(len(data_arrays)) :
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1070 self._datas.append( numpy.asmatrix(data_arrays[k]) )
319 6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1071 if self._datas[-1].shape[0] == 1 and self._datas[-1].shape[1] == num_examples :
6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1072 self._datas[-1] = self._datas[-1].transpose()
6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1073 for k in range(len(self._datas)) :
6441e568745e bug fixed when one matrix is an array, a 1-d matrix Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 318 diff changeset	1074 assert self._datas[k].shape[0] == num_examples
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1075
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1076
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1077 def __len__(self) :
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1078 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1079 Length of the dataset is based on the first array = data_arrays[0], using its shape
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1080 """
318 e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1081 return self._datas[0].shape[0]
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1082
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1083 def fieldNames(self) :
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1084 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1085 Returns the fieldnames as set in self.__init__
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1086 """
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1087 return self._fieldnames
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1088
318 e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1089 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1090 cursor = Example(fieldnames,[0]*len(fieldnames))
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1091 fieldnames = self.fieldNames() if fieldnames is None else fieldnames
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1092 for n in xrange(n_batches):
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1093 if offset == len(self):
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1094 break
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1095 for f in range(len(cursor._names)) :
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1096 idx = self._fieldnames.index(cursor._names[f])
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1097 assert idx >= 0
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1098 sub_data = self._datas[f][offset : offset+minibatch_size]
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1099 cursor._values[f] = sub_data
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1100 offset += len(sub_data) #can be less than minibatch_size at end
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1101 yield cursor
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1102
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1103 #return ArrayDataSetIterator(self,fieldnames,minibatch_size,n_batches,offset)
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1104
e2eab74b6a28 NArraysDataSet, a generalization ArrayDataSet where every field is a ndarray, is implemented. Not really tested aside basic stuff... Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 317 diff changeset	1105
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1106
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1107
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1108 class ArrayDataSet(ArrayFieldsDataSet):
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1109 """
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1110 An ArrayDataSet stores the fields as groups of columns in a numpy tensor,
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1111 whose first axis iterates over examples, second axis determines fields.
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1112 If the underlying array is N-dimensional (has N axes), then the field
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1113 values are (N-2)-dimensional objects (i.e. ordinary numbers if N=2).
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1114 """
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1115
188 f01ac276c6fb added __contains__ to Dataset, added parent constructor call to ArrayDataSet James Bergstra <bergstrj@iro.umontreal.ca> parents: 171 diff changeset	1116 def __init__(self, data_array, fields_columns, **kwargs):
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1117 """
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1118 Construct an ArrayDataSet from the underlying numpy array (data) and
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1119 a map (fields_columns) from fieldnames to field columns. The columns of a field are specified
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1120 using the standard arguments for indexing/slicing: integer for a column index,
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1121 slice for an interval of columns (with possible stride), or iterable of column indices.
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1122 """
188 f01ac276c6fb added __contains__ to Dataset, added parent constructor call to ArrayDataSet James Bergstra <bergstrj@iro.umontreal.ca> parents: 171 diff changeset	1123 ArrayFieldsDataSet.__init__(self, **kwargs)
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1124 self.data=data_array
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1125 self.fields_columns=fields_columns
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1126
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1127 # check consistency and complete slices definitions
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1128 for fieldname, fieldcolumns in self.fields_columns.items():
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1129 if type(fieldcolumns) is int:
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1130 assert fieldcolumns>=0 and fieldcolumns<data_array.shape[1]
268 3f1cd8897fda reverting dataset James Bergstra <bergstrj@iro.umontreal.ca> parents: 266 diff changeset	1131 if 1:
227 17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1132 #I changed this because it didn't make sense to me,
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1133 # and it made it more difficult to write my learner.
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1134 # If it breaks stuff, let's talk about it.
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1135 # - James 22/05/2008
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1136 self.fields_columns[fieldname]=[fieldcolumns]
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1137 else:
17c5d080964b reinstating changeset 216 clobbered accidentally by 218 James Bergstra <bergstrj@iro.umontreal.ca> parents: 223 diff changeset	1138 self.fields_columns[fieldname]=fieldcolumns
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1139 elif type(fieldcolumns) is slice:
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1140 start,step=None,None
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1141 if not fieldcolumns.start:
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1142 start=0
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1143 if not fieldcolumns.step:
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1144 step=1
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1145 if start or step:
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1146 self.fields_columns[fieldname]=slice(start,fieldcolumns.stop,step)
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1147 elif hasattr(fieldcolumns,"__iter__"): # something like a list
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1148 for i in fieldcolumns:
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1149 assert i>=0 and i<data_array.shape[1]
283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1150
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1151 def fieldNames(self):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1152 return self.fields_columns.keys()
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1153
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1154 def __len__(self):
9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1155 return len(self.data)
41 283e95c15b47 Added ArrayDataSet bengioy@grenat.iro.umontreal.ca parents: 40 diff changeset	1156
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1157 def __getitem__(self,key):
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1158 """More efficient implementation than the default __getitem__"""
66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1159 fieldnames=self.fields_columns.keys()
243 c8f19a9eb10f Optimisation in ArrayDataSet::__getitem__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 242 diff changeset	1160 values=self.fields_columns.values()
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1161 if type(key) is int:
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1162 return Example(fieldnames,
243 c8f19a9eb10f Optimisation in ArrayDataSet::__getitem__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 242 diff changeset	1163 [self.data[key,col] for col in values])
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1164 if type(key) is slice:
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1165 return Example(fieldnames,[self.data[key,col] for col in values])
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1166 if type(key) is list:
40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1167 for i in range(len(key)):
40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1168 if self.hasFields(key[i]):
40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1169 key[i]=self.fields_columns[key[i]]
316 5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1170 return Example(fieldnames,
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1171 #we must separate differently for list as numpy
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1172 # doesn't support self.data[[i1,...],[i2,...]]
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1173 # when their is more then two i1 and i2
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1174 [self.data[key,:][:,col]
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1175 if isinstance(col,list) else
5fe6d0c93109 getitem in ArrayDataSet is set up again, supposed to be faster than default one, has been tested agains the default behaviour. In particular, now always return a LookupList Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 314 diff changeset	1176 self.data[key,col] for col in values])
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1177
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1178 # else check for a fieldname
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1179 if self.hasFields(key):
105 8c0a1b11b007 bugfix, we keep all the line, but only a some columns Frederic Bastien <bastienf@iro.umontreal.ca> parents: 101 diff changeset	1180 return self.data[:,self.fields_columns[key]]
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1181 # else we are trying to access a property of the dataset
80 40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1182 assert key in self.__dict__ # else it means we are trying to access a non-existing property
40476a7746e8 bugfix Frederic Bastien <bastienf@iro.umontreal.ca> parents: 66 diff changeset	1183 return self.__dict__[key]
55 66619ce44497 Efficient implementation of getitem for ArrayDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 48 diff changeset	1184
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1185 def dontuse__iter__(self):
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1186 class ArrayDataSetIteratorIter(object):
1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1187 def __init__(self,dataset,fieldnames):
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1188 if fieldnames is None: fieldnames = dataset.fieldNames()
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1189 # store the resulting minibatch in a lookup-list of values
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1190 self.minibatch = Example(fieldnames,[0]*len(fieldnames))
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1191 self.dataset=dataset
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1192 self.current=0
238 ae1d85aca858 optimization in ArrayDataSet::__iter__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 235 diff changeset	1193 self.columns = [self.dataset.fields_columns[f]
ae1d85aca858 optimization in ArrayDataSet::__iter__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 235 diff changeset	1194 for f in self.minibatch._names]
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1195 self.l = self.dataset.data.shape[0]
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1196 def __iter__(self):
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1197 return self
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1198 def next(self):
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1199 #@todo: we suppose that we need to stop only when minibatch_size == 1.
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1200 # Otherwise, MinibatchWrapAroundIterator do it.
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1201 if self.current>=self.l:
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1202 raise StopIteration
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1203 sub_data = self.dataset.data[self.current]
238 ae1d85aca858 optimization in ArrayDataSet::__iter__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 235 diff changeset	1204 self.minibatch._values = [sub_data[c] for c in self.columns]
ae1d85aca858 optimization in ArrayDataSet::__iter__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 235 diff changeset	1205
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1206 self.current+=1
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1207 return self.minibatch
6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1208
270 1cafd495098c code cleanup and small optimisation Frederic Bastien <bastienf@iro.umontreal.ca> parents: 254 diff changeset	1209 return ArrayDataSetIteratorIter(self,self.fieldNames())
228 6f55e301c687 optimisation of ArrayDataSet Frederic Bastien <bastienf@iro.umontreal.ca> parents: 203 diff changeset	1210
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1211 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1212 cursor = Example(fieldnames,[0]*len(fieldnames))
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1213 fieldnames = self.fieldNames() if fieldnames is None else fieldnames
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1214 for n in xrange(n_batches):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1215 if offset == len(self):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1216 break
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1217 sub_data = self.data[offset : offset+minibatch_size]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1218 offset += len(sub_data) #can be less than minibatch_size at end
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1219 cursor._values = [sub_data[:,self.fields_columns[f]] for f in cursor._names]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1220 yield cursor
42 9b68774fcc6b Testing basic functionality and removing obvious bugs bengioy@grenat.iro.umontreal.ca parents: 41 diff changeset	1221
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1222 #return ArrayDataSetIterator(self,fieldnames,minibatch_size,n_batches,offset)
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1223
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1224
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1225 class CachedDataSet(DataSet):
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1226 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1227 Wrap a L{DataSet} whose values are computationally expensive to obtain
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1228 (e.g. because they involve some computation, or disk access),
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1229 so that repeated accesses to the same example are done cheaply,
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1230 by caching every example value that has been accessed at least once.
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1231
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1232 Optionally, for finite-length dataset, all the values can be computed
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1233 (and cached) upon construction of the CachedDataSet, rather at the
1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1234 first access.
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1235
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1236 @todo: when cache_all_upon_construction create mini-batches that are as
77 1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1237 large as possible but not so large as to fill up memory.
1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1238
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1239 @todo: add disk-buffering capability, so that when the cache becomes too
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1240 big for memory, we cache things on disk, trying to keep in memory only
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1241 the record most likely to be accessed next.
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1242 """
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1243 def __init__(self,source_dataset,cache_all_upon_construction=False):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1244 self.source_dataset=source_dataset
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1245 self.cache_all_upon_construction=cache_all_upon_construction
152 3f627e844cba Fixes in CacheDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 151 diff changeset	1246 self.cached_examples = []
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1247 if cache_all_upon_construction:
77 1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1248 # this potentially brings all the source examples
1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1249 # into memory at once, which may be too much
1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1250 # the work could possibly be done by minibatches
1e2bb5bad636 toying with different ways to implement learners bengioy@bengiomac.local parents: 74 diff changeset	1251 # that are as large as possible but no more than what memory allows.
146 8173e196e291 Trying to make CacheDataSet work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 144 diff changeset	1252 fields_values = source_dataset.minibatches(minibatch_size=len(source_dataset)).__iter__().next()
152 3f627e844cba Fixes in CacheDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 151 diff changeset	1253 assert all([len(self)==len(field_values) for field_values in fields_values])
3f627e844cba Fixes in CacheDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 151 diff changeset	1254 for example in fields_values.examples():
171 895b4b60f5e8 bugfix. Otherwise the example was writed over and not a new one was returned Frederic Bastien <bastienf@iro.umontreal.ca> parents: 167 diff changeset	1255 self.cached_examples.append(copy.copy(example))
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1256
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1257 self.fieldNames = source_dataset.fieldNames
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1258 self.hasFields = source_dataset.hasFields
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1259 self.valuesHStack = source_dataset.valuesHStack
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1260 self.valuesVStack = source_dataset.valuesVStack
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1261
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1262 def __len__(self):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1263 return len(self.source_dataset)
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1264
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1265 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1266 class CacheIterator(object):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1267 def __init__(self,dataset):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1268 self.dataset=dataset
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1269 self.current=offset
254 8ec867d12428 optimication in CachedDataSet.minibatches_nowrap Frederic Bastien <bastienf@iro.umontreal.ca> parents: 253 diff changeset	1270 self.all_fields = self.dataset.fieldNames()==fieldnames
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1271 def __iter__(self): return self
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1272 def next(self):
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1273 upper = self.current+minibatch_size
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1274 cache_len = len(self.dataset.cached_examples)
135 0d8e721cc63c Fixed bugs in dataset to make test_mlp.py work Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 134 diff changeset	1275 if upper>cache_len: # whole minibatch is not already in cache
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1276 # cache everything from current length to upper
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1277 for example in self.dataset.source_dataset[cache_len:upper]:
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1278 self.dataset.cached_examples.append(example)
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1279 all_fields_minibatch = Example(self.dataset.fieldNames(),
152 3f627e844cba Fixes in CacheDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 151 diff changeset	1280 zip(*self.dataset.cached_examples[self.current:self.current+minibatch_size]))
163 d7d67651d67c bugfix, we should advence by the minibatch size. Frederic Bastien <bastienf@iro.umontreal.ca> parents: 159 diff changeset	1281 self.current+=minibatch_size
254 8ec867d12428 optimication in CachedDataSet.minibatches_nowrap Frederic Bastien <bastienf@iro.umontreal.ca> parents: 253 diff changeset	1282 if self.all_fields:
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1283 return all_fields_minibatch
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1284 return Example(fieldnames,[all_fields_minibatch[name] for name in fieldnames])
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1285 return CacheIterator(self)
69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1286
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1287 def dontuse__getitem__(self,i):
153 71107b0ac860 optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 151 diff changeset	1288 if type(i)==int and len(self.cached_examples)>i:
71107b0ac860 optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 151 diff changeset	1289 return self.cached_examples[i]
71107b0ac860 optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 151 diff changeset	1290 else:
251 7e6edee187e3 optimization of CachedDataSet__getitem__ Frederic Bastien <bastienf@iro.umontreal.ca> parents: 243 diff changeset	1291 return self.source_dataset[i]
252 856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1292
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1293 def __iter__(self):
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1294 class CacheIteratorIter(object):
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1295 def __init__(self,dataset):
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1296 self.dataset=dataset
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1297 self.l = len(dataset)
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1298 self.current = 0
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1299 self.fieldnames = self.dataset.fieldNames()
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1300 self.example = Example(self.fieldnames,[0]*len(self.fieldnames))
252 856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1301 def __iter__(self): return self
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1302 def next(self):
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1303 if self.current>=self.l:
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1304 raise StopIteration
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1305 cache_len = len(self.dataset.cached_examples)
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1306 if self.current>=cache_len: # whole minibatch is not already in cache
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1307 # cache everything from current length to upper
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1308 self.dataset.cached_examples.append(
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1309 self.dataset.source_dataset[self.current])
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1310 self.example._values = self.dataset.cached_examples[self.current]
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1311 self.current+=1
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1312 return self.example
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1313
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1314 return CacheIteratorIter(self)
856d14dc4468 implemented CachedDataSet.__iter__ as an optimization Frederic Bastien <bastienf@iro.umontreal.ca> parents: 251 diff changeset	1315
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1316 class ApplyFunctionDataSet(DataSet):
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1317 """
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1318 A L{DataSet} that contains as fields the results of applying a
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1319 given function example-wise or minibatch-wise to all the fields of
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1320 an input dataset. The output of the function should be an iterable
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1321 (e.g. a list or a LookupList) over the resulting values.
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1322
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1323 The function take as input the fields of the dataset, not the examples.
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1324
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1325 In minibatch mode, the function is expected to work on minibatches
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1326 (takes a minibatch in input and returns a minibatch in output). More
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1327 precisely, it means that each element of the input or output list
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1328 should be iterable and indexable over the individual example values
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1329 (typically these elements will be numpy arrays). All of the elements
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1330 in the input and output lists should have the same length, which is
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1331 the length of the minibatch.
57 1aabd2e2bb5f Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet Yoshua Bengio <bengioy@iro.umontreal.ca> parents: 56 diff changeset	1332
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1333 The function is applied each time an example or a minibatch is accessed.
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1334 To avoid re-doing computation, wrap this dataset inside a CachedDataSet.
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1335
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1336 If the values_{h,v}stack functions are not provided, then
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1337 the input_dataset.values{H,V}Stack functions are used by default.
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1338
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1339 """
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1340
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1341 def __init__(self,input_dataset,function,output_names,minibatch_mode=True,
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1342 values_hstack=None,values_vstack=None,
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1343 description=None,fieldtypes=None):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1344 """
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1345 Constructor takes an input dataset that has as many fields as the function
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1346 expects as inputs. The resulting dataset has as many fields as the function
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1347 produces as outputs, and that should correspond to the number of output names
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1348 (provided in a list).
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1349
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1350 Note that the expected semantics of the function differs in minibatch mode
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1351 (it takes minibatches of inputs and produces minibatches of outputs, as
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1352 documented in the class comment).
211 bd728c83faff in __get__, problem if the i.stop was None, i being the slice, added one line replacing None by the len(self) Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca> parents: 203 diff changeset	1353
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1354 TBM: are filedtypes the old field types (from input_dataset) or the new ones
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1355 (for the new dataset created)?
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1356 """
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1357 self.input_dataset=input_dataset
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1358 self.function=function
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1359 self.output_names=output_names
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1360 self.minibatch_mode=minibatch_mode
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1361 DataSet.__init__(self,description,fieldtypes)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1362 self.valuesHStack = values_hstack if values_hstack else input_dataset.valuesHStack
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1363 self.valuesVStack = values_vstack if values_vstack else input_dataset.valuesVStack
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1364
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1365 def __len__(self):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1366 return len(self.input_dataset)
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1367
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1368 def fieldNames(self):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1369 return self.output_names
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1370
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1371 def minibatches_nowrap(self, fieldnames, args, *kwargs):
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1372 all_input_fieldNames = self.input_dataset.fieldNames()
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1373 mbnw = self.input_dataset.minibatches_nowrap
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1374
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1375 for input_fields in mbnw(all_input_fieldNames, args, *kwargs):
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1376 if self.minibatch_mode:
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1377 all_output_fields = self.function(*input_fields)
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1378 else:
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1379 input_examples = zip(*input_fields) #makes so that [i] means example i
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1380 output_examples = [self.function(*input_example)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1381 for input_example in input_examples]
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1382 all_output_fields = zip(*output_examples)
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1383
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1384 all_outputs = Example(self.output_names, all_output_fields)
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1385 #print 'input_fields', input_fields
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1386 #print 'all_outputs', all_outputs
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1387 if fieldnames==self.output_names:
293 4bfdda107a17 still merging James Bergstra <bergstrj@iro.umontreal.ca> parents: 292 diff changeset	1388 rval = all_outputs
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1389 else:
293 4bfdda107a17 still merging James Bergstra <bergstrj@iro.umontreal.ca> parents: 292 diff changeset	1390 rval = Example(fieldnames,[all_outputs[name] for name in fieldnames])
296 f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1391 #print 'rval', rval
f5d33f9c0b9c ApplyFunctionDataSet passing James Bergstra <bergstrj@iro.umontreal.ca> parents: 293 diff changeset	1392 #print '--------'
293 4bfdda107a17 still merging James Bergstra <bergstrj@iro.umontreal.ca> parents: 292 diff changeset	1393 yield rval
73 69f97aad3faf Coded untested ApplyFunctionDataSet and CacheDataSet bengioy@bengiomac.local parents: 72 diff changeset	1394
290 9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1395 def untested__iter__(self): # only implemented for increased efficiency
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1396 class ApplyFunctionSingleExampleIterator(object):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1397 def __init__(self,output_dataset):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1398 self.current=0
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1399 self.output_dataset=output_dataset
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1400 self.input_iterator=output_dataset.input_dataset.__iter__()
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1401 def __iter__(self): return self
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1402 def next(self):
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1403 if self.output_dataset.minibatch_mode:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1404 function_inputs = [[input] for input in self.input_iterator.next()]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1405 outputs = self.output_dataset.function(*function_inputs)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1406 assert all([hasattr(output,'__iter__') for output in outputs])
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1407 function_outputs = [output[0] for output in outputs]
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1408 else:
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1409 function_inputs = self.input_iterator.next()
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1410 function_outputs = self.output_dataset.function(*function_inputs)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1411 return Example(self.output_dataset.output_names,function_outputs)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1412 return ApplyFunctionSingleExampleIterator(self)
9b533cc7874a trying to get default implemenations to work James Bergstra <bergstrj@iro.umontreal.ca> parents: 274 diff changeset	1413
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1414 def supervised_learning_dataset(src_dataset,input_fields,target_fields,weight_field=None):
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1415 """
167 4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1416 Wraps an arbitrary L{DataSet} into one for supervised learning tasks
4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1417 by forcing the user to define a set of fields as the 'input' field
4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1418 and a set of fields as the 'target' field. Optionally, a single
4803cb76e26b Updated documentation Joseph Turian <turian@gmail.com> parents: 166 diff changeset	1419 weight_field can also be defined.
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1420 """
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1421 args = ((input_fields,'input'),(output_fields,'target'))
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1422 if weight_field: args+=(([weight_field],'weight'))
36 438440ba0627 Rewriting dataset.py completely bengioy@zircon.iro.umontreal.ca parents: 29 diff changeset	1423 return src_dataset.merge_fields(*args)
23 526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1424
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1425
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1426
526e192b0699 Working on ApplyFunctionDataSet, added constraint that bengioy@esprit.iro.umontreal.ca parents: 22 diff changeset	1427

Mercurial > pylearn

annotate dataset.py @ 320:b1da46b9b901