Mercurial > pylearn
annotate dataset.py @ 306:e2e5157ff044
added src_version cache
| author | James Bergstra <bergstrj@iro.umontreal.ca> |
|---|---|
| date | Tue, 10 Jun 2008 13:02:04 -0400 |
| parents | f5d33f9c0b9c |
| children | 923de30457f0 |
| 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 | 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 | 41 """A virtual base class for datasets. |
| 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 | 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 | 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 | 56 |
| 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 | 91 Note: Fields are not mutually exclusive, i.e. two fields can overlap in their actual content. |
| 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 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
102 - dataset[:n] returns a dataset with the n first examples. |
| 16 | 103 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
104 - dataset[i1:i2:s] returns a dataset 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 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
108 - dataset[[i1,i2,...in]] returns a dataset with examples i1,i2,...in. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
109 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
110 - dataset.<property> returns the value of a property associated with |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
111 the name <property>. The following properties should be supported: |
| 41 | 112 - '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
|
113 - 'fieldtypes': a list of types (one per field) |
| 78 | 114 A DataSet may have other attributes that it makes visible to other objects. These are |
| 115 used to store information that is not example-wise but global to the dataset. | |
| 116 The list of names of these attributes is given by the attribute_names() method. | |
| 41 | 117 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
118 Datasets can be concatenated either vertically (increasing the length) or |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
119 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
|
120 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
|
121 and numpy.vstack): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
122 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
123 - 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
|
124 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
125 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
|
126 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
|
127 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
128 - dataset1 & dataset2 & dataset3 == dataset.vstack([dataset1,dataset2,dataset3]) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
129 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
130 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
|
131 (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
|
132 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
|
133 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
134 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
|
135 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
|
136 a DataSetFields fields1 and fields2, and fields1 | fields2 concatenates their |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
137 examples. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
138 |
| 41 | 139 A dataset can hold arbitrary key-value pairs that may be used to access meta-data |
| 140 or other properties of the dataset or associated with the dataset or the result | |
| 141 of a computation stored in a dataset. These can be accessed through the [key] syntax | |
| 142 when key is a string (or more specifically, neither an integer, a slice, nor a list). | |
| 78 | 143 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
144 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
|
145 - __len__ if it is not a stream |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
146 - fieldNames |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
147 - minibatches_nowrap (called by DataSet.minibatches()) |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
148 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
|
149 - valuesHStack |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
150 - valuesVStack |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
151 - hasFields |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
152 - __getitem__ may not be feasible with some streams |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
153 - __iter__ |
| 78 | 154 A sub-class should also append attributes to self._attribute_names |
| 155 (the default value returned by attributeNames()). | |
| 156 By convention, attributes not in attributeNames() should have a name | |
| 157 starting with an underscore. | |
| 158 @todo enforce/test that convention! | |
|
2
3fddb1c8f955
Rewrote DataSet interface and created FiniteDataSet interface.
bengioy@bengiomac.local
parents:
1
diff
changeset
|
159 """ |
|
1
2cd82666b9a7
Added statscollector and started writing dataset and learner.
bengioy@esprit.iro.umontreal.ca
parents:
0
diff
changeset
|
160 |
| 83 | 161 numpy_vstack = lambda fieldname,values: numpy.vstack(values) |
| 162 numpy_hstack = lambda fieldnames,values: numpy.hstack(values) | |
|
77
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
163 |
| 292 | 164 def __init__(self, description=None, fieldnames=None, fieldtypes=None): |
| 165 """ | |
| 166 @type fieldnames: list of strings | |
| 167 @type fieldtypes: list of python types, same length as fieldnames | |
| 168 @type description: string | |
| 169 @param description: description/name for this dataset | |
| 170 """ | |
| 171 def default_desc(): | |
| 172 return type(self).__name__ \ | |
| 173 + " ( " + join([x.__name__ for x in type(self).__bases__]) + " )" | |
| 174 | |
| 175 #self.fieldnames = fieldnames | |
| 176 | |
| 177 self.fieldtypes = fieldtypes if fieldtypes is not None \ | |
| 178 else [None]*1 #len(fieldnames) | |
| 179 | |
| 180 self.description = default_desc() if description is None \ | |
| 181 else description | |
| 78 | 182 self._attribute_names = ["description"] |
| 292 | 183 |
| 184 attributeNames = property(lambda self: copy.copy(self._attribute_names)) | |
| 185 | |
| 186 def __contains__(self, fieldname): | |
| 187 return (fieldname in self.fieldNames()) \ | |
| 188 or (fieldname in self.attributeNames()) | |
| 189 | |
| 190 def __iter__(self): | |
| 191 """Supports the syntax "for i in dataset: ..." | |
| 78 | 192 |
| 292 | 193 Using this syntax, "i" will be an Example instance (or equivalent) with |
| 194 all the fields of DataSet self. Every field of "i" will give access to | |
| 195 a field of a single example. Fields should be accessible via | |
| 196 i["fielname"] or i[3] (in the order defined by the elements of the | |
| 197 Example returned by this iterator), but the derived class is free | |
| 198 to accept any type of identifier, and add extra functionality to the iterator. | |
| 199 | |
| 200 The default implementation calls the minibatches iterator and extracts the first example of each field. | |
| 201 """ | |
| 202 return DataSet.MinibatchToSingleExampleIterator(self.minibatches(None, minibatch_size = 1)) | |
| 203 | |
| 204 def __len__(self): | |
| 205 """ | |
| 206 len(dataset) returns the number of examples in the dataset. | |
| 207 By default, a DataSet is a 'stream', i.e. it has an unbounded length (sys.maxint). | |
| 208 Sub-classes which implement finite-length datasets should redefine this method. | |
| 209 Some methods only make sense for finite-length datasets. | |
| 210 """ | |
| 211 return None | |
| 212 | |
| 78 | 213 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
214 class MinibatchToSingleExampleIterator(object): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
215 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
216 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
|
217 single-example values in the result. Therefore the result of |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
218 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
|
219 (whereas the result of iterating over minibatches gives in each |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
220 Example field an iterable object over the individual examples in |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
221 the minibatch). |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
222 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
223 def __init__(self, minibatch_iterator): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
224 self.minibatch_iterator = minibatch_iterator |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
225 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
|
226 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
|
227 return self |
|
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
228 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
|
229 size1_minibatch = self.minibatch_iterator.next() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
230 if not self.minibatch: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
231 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
|
232 else: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
233 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
|
234 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
|
235 |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
236 def next_index(self): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
237 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
|
238 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
239 class MinibatchWrapAroundIterator(object): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
240 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
241 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
|
242 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
|
243 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
|
244 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
|
245 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
|
246 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
|
247 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
248 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
|
249 self.dataset=dataset |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
250 self.fieldnames=fieldnames |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
251 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
|
252 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
|
253 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
|
254 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
|
255 self.L=len(dataset) |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
256 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
|
257 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
|
258 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
|
259 ds_nbatches = min(n_batches,ds_nbatches) |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
260 if fieldnames: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
261 assert dataset.hasFields(*fieldnames) |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
262 else: |
|
98
7186e4f502d1
bugfix in DataSet.minibatch to correctly wrap in all corner case.
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
95
diff
changeset
|
263 self.fieldnames=dataset.fieldNames() |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
264 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
|
265 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
266 def __iter__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
267 return self |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
268 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
269 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
|
270 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
|
271 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
272 def next(self): |
|
43
e92244f30116
Corrected iterator logic errors
bengioy@grenat.iro.umontreal.ca
parents:
42
diff
changeset
|
273 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
|
274 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
|
275 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
|
276 raise StopIteration |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
277 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
|
278 if upper <=self.L: |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
279 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
|
280 else: |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
281 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
|
282 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
|
283 # 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
|
284 # 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
|
285 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
|
286 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
|
287 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
|
288 # 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
|
289 # 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
|
290 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
|
291 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
|
292 minibatch = Example(self.fieldnames, |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
293 [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
|
294 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
|
295 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
|
296 self.n_batches_done+=1 |
|
43
e92244f30116
Corrected iterator logic errors
bengioy@grenat.iro.umontreal.ca
parents:
42
diff
changeset
|
297 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
|
298 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
|
299 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
|
300 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
|
301 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
|
302 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
|
303 ds_nbatches,self.next_row) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
304 return DataSetFields(MinibatchDataSet(minibatch,self.dataset.valuesVStack, |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
305 self.dataset.valuesHStack), |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
306 minibatch.keys()) |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
307 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
308 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
309 minibatches_fieldnames = None |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
310 minibatches_minibatch_size = 1 |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
311 minibatches_n_batches = None |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
312 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
|
313 fieldnames = minibatches_fieldnames, |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
314 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
|
315 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
|
316 offset = 0): |
|
6
d5738b79089a
Removed MinibatchIterator and instead made minibatch_size a field of all DataSets,
bengioy@bengiomac.local
parents:
5
diff
changeset
|
317 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
318 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
|
319 |
|
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
320 for i in dataset.minibatches(None,**kwargs): ... |
| 16 | 321 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
322 for i in dataset.minibatches([f1, f2, f3],**kwargs): ... |
| 16 | 323 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
324 for i1, i2, i3 in dataset.minibatches([f1, f2, f3],**kwargs): ... |
| 16 | 325 |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
326 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
|
327 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
|
328 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
diff
changeset
|
329 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
diff
changeset
|
330 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
|
331 |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
332 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
|
333 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
diff
changeset
|
334 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
|
335 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
336 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
|
337 a DataSetFields object, which is a Example (indexed by the field names) whose |
| 80 | 338 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
|
339 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
|
340 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
|
341 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
|
342 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
343 PARAMETERS |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
344 - fieldnames (list of any type, default None): |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
345 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
diff
changeset
|
346 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
diff
changeset
|
347 derived class can choose a default, e.g. all fields. |
| 16 | 348 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
349 - minibatch_size (integer, default 1) |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
350 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
diff
changeset
|
351 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
diff
changeset
|
352 |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
353 - n_batches (integer, default None) |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
354 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
diff
changeset
|
355 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
diff
changeset
|
356 iterator should support looping indefinitely. |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
357 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
358 - 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
|
359 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
|
360 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
361 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
diff
changeset
|
362 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
diff
changeset
|
363 |
|
11
be128b9127c8
Debugged (to the extent of my tests) the new version of dataset
bengioy@esprit.iro.umontreal.ca
parents:
9
diff
changeset
|
364 """ |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
365 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
|
366 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
367 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
|
368 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
369 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
|
370 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
|
371 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
|
372 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
|
373 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
|
374 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
|
375 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
|
376 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
|
377 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
|
378 """ |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
379 raise AbstractFunction() |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
380 |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
381 def is_unbounded(self): |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
382 """ |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
383 Tests whether a dataset is unbounded (e.g. a stream). |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
384 """ |
| 123 | 385 return len(self)==maxint |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
386 |
|
26
672fe4b23032
Fixed dataset errors so that _test_dataset.py works again.
bengioy@grenat.iro.umontreal.ca
parents:
23
diff
changeset
|
387 def hasFields(self,*fieldnames): |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
388 """ |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
389 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
|
390 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
|
391 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
|
392 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
393 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
|
394 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
|
395 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
396 return len(unique_elements_list_intersection(fieldnames,self.fieldNames()))>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
397 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
398 def fieldNames(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
399 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
400 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
|
401 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
|
402 """ |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
403 raise AbstractFunction() |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
404 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
405 def __call__(self,*fieldnames): |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
406 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
407 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
|
408 """ |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
409 assert self.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
410 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
|
411 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
412 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
|
413 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
414 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
diff
changeset
|
415 """ |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
416 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
|
417 |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
418 def getitem_key(self, fieldname): |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
419 """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
|
420 getitem. |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
421 """ |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
422 #removing as per discussion June 4. --JSB |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
423 |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
424 i = fieldname |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
425 # else check for a fieldname |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
426 if self.hasFields(i): |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
427 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
|
428 # 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
|
429 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
|
430 return self.__dict__[i] |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
431 |
|
2
3fddb1c8f955
Rewrote DataSet interface and created FiniteDataSet interface.
bengioy@bengiomac.local
parents:
1
diff
changeset
|
432 def __getitem__(self,i): |
|
28
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
433 """ |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
434 @rtype: Example |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
435 @returns: single or multiple examples |
|
1
2cd82666b9a7
Added statscollector and started writing dataset and learner.
bengioy@esprit.iro.umontreal.ca
parents:
0
diff
changeset
|
436 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
437 @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
|
438 @param i: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
439 dataset[i] returns the (i+1)-th example of the dataset. |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
440 dataset[i:j] returns the subdataset with examples i,i+1,...,j-1. |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
441 dataset[i:j:s] returns the subdataset with examples i,i+2,i+4...,j-2. |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
442 dataset[[i1,i2,..,in]] returns the subdataset with examples i1,i2,...,in. |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
443 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
444 @note: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
445 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
|
446 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
|
447 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
|
448 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
|
449 |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
450 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
|
451 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
|
452 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
|
453 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
|
454 """ |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
455 |
|
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
|
456 if type(i) is int: |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
457 #TODO: consider asserting that i >= 0 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
458 i_batch = self.minibatches_nowrap(self.fieldNames(), |
| 293 | 459 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
|
460 return DataSet.MinibatchToSingleExampleIterator(i_batch).next() |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
461 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
462 #if i is a contiguous slice |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
463 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
|
464 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
|
465 upper_bound = len(self) if i.stop is None else i.stop |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
466 return MinibatchDataSet(self.minibatches_nowrap(self.fieldNames(), |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
467 minibatch_size=upper_bound - offset, |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
468 n_batches=1, |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
469 offset=offset).next()) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
470 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
471 # 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
|
472 # 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
|
473 if type(i) is slice: |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
474 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
|
475 upper_bound = len(self) if i.stop is None else i.stop |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
476 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
|
477 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
478 # handle tuples, arrays, lists |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
479 if hasattr(i, '__getitem__'): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
480 for idx in i: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
481 #dis-allow nested slices |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
482 if not isinstance(idx, int): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
483 raise TypeError(idx) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
484 # call back into self.__getitem__ |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
485 examples = [self.minibatches_nowrap(self.fieldNames(), |
| 293 | 486 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
|
487 for ii in i] |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
488 # 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
|
489 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
|
490 for e in examples: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
491 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
|
492 f.append(v) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
493 #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
|
494 zz = zip(self.fieldNames(),field_values) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
495 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
|
496 example = Example(self.fieldNames(), vst) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
497 return MinibatchDataSet(example, self.valuesVStack, self.valuesHStack) |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
498 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
|
499 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
500 def valuesHStack(self,fieldnames,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
501 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
502 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
|
503 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
|
504 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
|
505 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
|
506 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
|
507 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
|
508 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
|
509 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
|
510 pointing to the original values for other data types. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
511 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
512 all_numpy=True |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
513 for value in fieldvalues: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
514 if not type(value) is numpy.ndarray: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
515 all_numpy=False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
516 if all_numpy: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
517 return numpy.hstack(fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
518 # 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
|
519 return fieldvalues |
|
26
672fe4b23032
Fixed dataset errors so that _test_dataset.py works again.
bengioy@grenat.iro.umontreal.ca
parents:
23
diff
changeset
|
520 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
521 def valuesVStack(self,fieldname,values): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
522 """ |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
523 @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
|
524 @type fieldname: any type |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
525 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
526 @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
|
527 @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
|
528 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
529 @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
|
530 @rtype: something suitable as a minibatch field |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
531 """ |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
532 rval = [] |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
533 for v in values: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
534 rval.extend(v) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
535 return rval |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
536 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
537 def __or__(self,other): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
538 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
539 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
|
540 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
|
541 """ |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
542 return HStackedDataSet([self,other]) |
|
3
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
543 |
|
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
|
544 def __and__(self,other): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
545 """ |
|
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
|
546 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
|
547 (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
|
548 works if they all have the same fields. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
549 """ |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
550 return VStackedDataSet([self,other]) |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
551 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
552 def hstack(datasets): |
|
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 hstack(dataset1,dataset2,...) returns dataset1 | datataset2 | ... |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
555 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
|
556 of the individual datasets. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
557 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
558 assert len(datasets)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
559 if len(datasets)==1: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
560 return datasets[0] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
561 return HStackedDataSet(datasets) |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
562 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
563 def vstack(datasets): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
564 """ |
|
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
|
565 vstack(dataset1,dataset2,...) returns dataset1 & datataset2 & ... |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
566 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
|
567 over those of dataset2, etc. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
568 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
569 assert len(datasets)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
570 if len(datasets)==1: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
571 return datasets[0] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
572 return VStackedDataSet(datasets) |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
573 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
574 class FieldsSubsetDataSet(DataSet): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
575 """ |
| 167 | 576 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
|
577 """ |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
578 def __init__(self,src,fieldnames): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
579 self.src=src |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
580 self.fieldnames=fieldnames |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
581 assert src.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
582 self.valuesHStack = src.valuesHStack |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
583 self.valuesVStack = src.valuesVStack |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
584 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
585 def __len__(self): return len(self.src) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
586 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
587 def fieldNames(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
588 return self.fieldnames |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
589 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
590 def __iter__(self): |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
591 class FieldsSubsetIterator(object): |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
592 def __init__(self,ds): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
593 self.ds=ds |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
594 self.src_iter=ds.src.__iter__() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
595 self.example=None |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
596 def __iter__(self): return self |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
597 def next(self): |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
598 complete_example = self.src_iter.next() |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
599 if self.example: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
600 self.example._values=[complete_example[field] |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
601 for field in self.ds.fieldnames] |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
602 else: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
603 self.example=Example(self.ds.fieldnames, |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
604 [complete_example[field] for field in self.ds.fieldnames]) |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
605 return self.example |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
606 return FieldsSubsetIterator(self) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
607 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
608 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
|
609 assert self.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
610 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
|
611 def dontuse__getitem__(self,i): |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
612 return FieldsSubsetDataSet(self.src[i],self.fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
613 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
614 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
615 class DataSetFields(Example): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
616 """ |
| 167 | 617 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
|
618 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
|
619 as a transpose of the associated dataset. |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
620 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
621 To iterate over fields, one can do |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
622 * for fields in dataset.fields() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
623 * 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
|
624 * 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
|
625 and each of these fields is iterable over the examples: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
626 * for field_examples in dataset.fields(): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
627 for example_value in field_examples: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
628 ... |
|
241
ddb88a8e9fd2
If I understand properly, the length of an unbounded stream is sys.maxint
delallea@opale.iro.umontreal.ca
parents:
231
diff
changeset
|
629 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
|
630 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
|
631 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
|
632 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
|
633 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
|
634 be turned back into a DataSet with its examples() method: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
635 dataset2 = dataset1.fields().examples() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
636 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
|
637 |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
638 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
|
639 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
|
640 the examples. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
641 """ |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
642 def __init__(self,dataset,fieldnames): |
|
65
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
643 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
|
644 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
|
645 fieldnames=dataset.fieldNames() |
|
274
ed70580f2324
bugfix in FieldSubsetDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
273
diff
changeset
|
646 elif not list(fieldnames)==list(dataset.fieldNames()): |
|
ed70580f2324
bugfix in FieldSubsetDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
273
diff
changeset
|
647 #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
|
648 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
|
649 assert dataset.hasFields(*fieldnames) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
650 self.dataset=dataset |
|
66
dde1fb1b63ba
fixed test and removed print
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
65
diff
changeset
|
651 |
|
64
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
652 if isinstance(dataset,MinibatchDataSet): |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
653 Example.__init__(self,fieldnames,list(dataset._fields)) |
|
65
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
654 elif isinstance(original_dataset,MinibatchDataSet): |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
655 Example.__init__(self,fieldnames, |
|
65
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
656 [original_dataset._fields[field] |
|
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
657 for field in fieldnames]) |
|
64
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
658 else: |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
659 minibatch_iterator = dataset.minibatches(fieldnames, |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
660 minibatch_size=len(dataset), |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
661 n_batches=1) |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
662 minibatch=minibatch_iterator.next() |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
663 Example.__init__(self,fieldnames,minibatch) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
664 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
665 def examples(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
666 return self.dataset |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
667 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
668 def __or__(self,other): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
669 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
670 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
|
671 of the list of examples of DataSetFields fields1 and fields2. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
672 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
673 return (self.examples() + other.examples()).fields() |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
674 |
|
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
|
675 def __and__(self,other): |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
676 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
677 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
|
678 of the fields of DataSetFields fields1 and fields2. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
679 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
680 return (self.examples() | other.examples()).fields() |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
681 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
682 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
683 class MinibatchDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
684 """ |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
685 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
|
686 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
|
687 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
688 def __init__(self,fields_lookuplist,values_vstack=DataSet().valuesVStack, |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
689 values_hstack=DataSet().valuesHStack): |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
690 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
691 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
|
692 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
|
693 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
|
694 """ |
|
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
|
695 |
|
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
|
696 self._fields=fields_lookuplist |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
697 assert len(fields_lookuplist)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
698 self.length=len(fields_lookuplist[0]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
699 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
|
700 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
|
701 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
|
702 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
|
703 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
|
704 print 'field=',field |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
705 print 'fields_lookuplist=', fields_lookuplist |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
706 assert self.length==len(field) |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
707 self.valuesVStack=values_vstack |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
708 self.valuesHStack=values_hstack |
|
3
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
709 |
|
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
710 def __len__(self): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
711 return self.length |
|
28
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
712 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
713 def dontuse__getitem__(self,i): |
| 80 | 714 if type(i) in (slice,list): |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
715 return DataSetFields(MinibatchDataSet( |
| 80 | 716 Example(self._fields.keys(),[field[i] for field in self._fields])),self.fieldNames()) |
| 717 if type(i) is int: | |
| 85 | 718 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
|
719 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
|
720 return self._fields[i] |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
721 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
|
722 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
|
723 |
|
29
46c5c90019c2
Changed apply_function so that it propagates methods of the source.
bengioy@grenat.iro.umontreal.ca
parents:
28
diff
changeset
|
724 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
|
725 return self._fields.keys() |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
726 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
727 def hasFields(self,*fieldnames): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
728 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
|
729 if fieldname not in self._fields.keys(): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
730 return False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
731 return True |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
732 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
733 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
|
734 #@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
|
735 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
|
736 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
|
737 # 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
|
738 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
|
739 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
|
740 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
741 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
|
742 self.next_example=offset |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
743 assert minibatch_size >= 0 |
| 41 | 744 if offset+minibatch_size > ds.length: |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
745 raise NotImplementedError() |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
746 def __iter__(self): |
|
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
747 return self |
|
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
748 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
|
749 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
|
750 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
|
751 #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
|
752 # [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
|
753 # 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
|
754 # 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
|
755 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
|
756 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
|
757 #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
|
758 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
|
759 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
|
760 #print minibatch |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
761 self.next_example+=minibatch_size |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
762 return minibatch |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
763 |
|
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
|
764 # 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
|
765 return Iterator(self,fieldnames) |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
766 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
767 class HStackedDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
768 """ |
| 167 | 769 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
|
770 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
|
771 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
772 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
|
773 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
|
774 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
|
775 |
| 167 | 776 @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
|
777 """ |
| 41 | 778 def __init__(self,datasets,accept_nonunique_names=False,description=None,field_types=None): |
| 779 DataSet.__init__(self,description,field_types) | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
780 self.datasets=datasets |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
781 self.accept_nonunique_names=accept_nonunique_names |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
782 self.fieldname2dataset={} |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
783 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
784 def rename_field(fieldname,dataset,i): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
785 if hasattr(dataset,"__name__"): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
786 return dataset.__name__ + "." + fieldname |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
787 return fieldname+"."+str(i) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
788 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
789 # 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
|
790 self.length=None |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
791 names_to_change=[] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
792 for i in xrange(len(datasets)): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
793 dataset = datasets[i] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
794 length=len(dataset) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
795 if self.length: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
796 assert self.length==length |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
797 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
798 self.length=length |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
799 for fieldname in dataset.fieldNames(): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
800 if fieldname in self.fieldname2dataset: # name conflict! |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
801 if accept_nonunique_names: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
802 fieldname=rename_field(fieldname,dataset,i) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
803 names2change.append((fieldname,i)) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
804 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
805 raise ValueError("Incompatible datasets: non-unique field name = "+fieldname) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
806 self.fieldname2dataset[fieldname]=i |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
807 for fieldname,i in names_to_change: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
808 del self.fieldname2dataset[fieldname] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
809 self.fieldname2dataset[rename_field(fieldname,self.datasets[i],i)]=i |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
810 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
811 def hasFields(self,*fieldnames): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
812 for fieldname in fieldnames: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
813 if not fieldname in self.fieldname2dataset: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
814 return False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
815 return True |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
816 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
817 def fieldNames(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
818 return self.fieldname2dataset.keys() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
819 |
| 41 | 820 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
|
821 |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
822 class HStackedIterator(object): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
823 def __init__(self,hsds,iterators): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
824 self.hsds=hsds |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
825 self.iterators=iterators |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
826 def __iter__(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
827 return self |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
828 def next(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
829 # 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
|
830 l=Example() |
| 140 | 831 for iter in self.iterators: |
| 832 l.append_lookuplist(iter.next()) | |
| 833 return l | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
834 |
| 125 | 835 assert self.hasFields(*fieldnames) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
836 # 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
|
837 # and construct corresponding minibatch iterators |
| 140 | 838 if fieldnames and fieldnames!=self.fieldNames(): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
839 datasets=set([]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
840 fields_in_dataset=dict([(dataset,[]) for dataset in datasets]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
841 for fieldname in fieldnames: |
| 136 | 842 dataset=self.datasets[self.fieldname2dataset[fieldname]] |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
843 datasets.add(dataset) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
844 fields_in_dataset[dataset].append(fieldname) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
845 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
|
846 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
|
847 for dataset in datasets] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
848 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
849 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
|
850 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
|
851 return HStackedIterator(self,iterators) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
852 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
853 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
854 def untested_valuesVStack(self,fieldname,fieldvalues): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
855 return self.datasets[self.fieldname2dataset[fieldname]].valuesVStack(fieldname,fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
856 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
857 def untested_valuesHStack(self,fieldnames,fieldvalues): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
858 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
859 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
|
860 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
|
861 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
|
862 all the fieldnames are of the same sub-dataset. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
863 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
864 return self.datasets[self.fieldname2dataset[fieldnames[0]]].valuesHStack(fieldnames,fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
865 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
866 class VStackedDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
867 """ |
| 167 | 868 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
|
869 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
|
870 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
|
871 |
| 167 | 872 @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
|
873 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
874 def __init__(self,datasets): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
875 self.datasets=datasets |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
876 self.length=0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
877 self.index2dataset={} |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
878 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
|
879 fieldnames = datasets[-1].fieldNames() |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
880 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
|
881 # 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
|
882 # 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
|
883 for dataset,k in enumerate(datasets[0:-1]): |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
884 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
|
885 L=len(dataset) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
886 for i in xrange(L): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
887 self.index2dataset[self.length+i]=k |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
888 self.datasets_start_row.append(self.length) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
889 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
|
890 assert dataset.fieldNames()==fieldnames |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
891 self.datasets_start_row.append(self.length) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
892 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
|
893 # 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
|
894 # 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
|
895 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
|
896 # 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
|
897 # TODO |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
898 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
|
899 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
900 def __len__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
901 return self.length |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
902 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
903 def fieldNames(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
904 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
|
905 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
906 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
|
907 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
|
908 |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
909 def locate_row(self,row): |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
910 """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
|
911 dataset_index = self.index2dataset[row] |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
912 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
|
913 return dataset_index, row_within_dataset |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
914 |
| 41 | 915 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
|
916 |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
917 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
|
918 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
|
919 self.vsds=vsds |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
920 self.next_row=offset |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
921 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
|
922 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
|
923 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
|
924 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
925 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
|
926 L=len(dataset) |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
927 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
|
928 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
|
929 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
|
930 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
|
931 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
|
932 n_left_in_mb=minibatch_size-L |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
933 ds_nbatches=1 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
934 else: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
935 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
|
936 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
|
937 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
|
938 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
939 def move_to_next_dataset(self): |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
940 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
|
941 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
|
942 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
|
943 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
|
944 else: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
945 self.next_dataset_index +=1 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
946 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
|
947 self.next_dataset_index = 0 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
948 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
|
949 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
|
950 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
951 def __iter__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
952 return self |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
953 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
954 def next(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
955 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
|
956 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
|
957 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
|
958 extra_mb = [] |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
959 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
|
960 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
|
961 extra_mb.append(self.next_iterator.next()) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
962 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
|
963 [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
|
964 [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
|
965 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
|
966 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
967 self.next_row+=minibatch_size |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
968 self.next_dataset_row+=minibatch_size |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
969 if self.next_row+minibatch_size>len(dataset): |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
970 self.move_to_next_dataset() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
971 return examples |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
972 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
|
973 |
| 41 | 974 class ArrayFieldsDataSet(DataSet): |
| 975 """ | |
| 976 Virtual super-class of datasets whose field values are numpy array, | |
| 977 thus defining valuesHStack and valuesVStack for sub-classes. | |
| 978 """ | |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
979 def __init__(self,description=None,field_types=None): |
|
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
980 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
|
981 def untested_valuesHStack(self,fieldnames,fieldvalues): |
| 41 | 982 """Concatenate field values horizontally, e.g. two vectors |
| 983 become a longer vector, two matrices become a wider matrix, etc.""" | |
| 984 return numpy.hstack(fieldvalues) | |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
985 def untested_valuesVStack(self,fieldname,values): |
| 41 | 986 """Concatenate field values vertically, e.g. two vectors |
| 987 become a two-row matrix, two matrices become a longer matrix, etc.""" | |
| 988 return numpy.vstack(values) | |
| 989 | |
| 990 class ArrayDataSet(ArrayFieldsDataSet): | |
| 991 """ | |
| 992 An ArrayDataSet stores the fields as groups of columns in a numpy tensor, | |
| 993 whose first axis iterates over examples, second axis determines fields. | |
| 994 If the underlying array is N-dimensional (has N axes), then the field | |
| 995 values are (N-2)-dimensional objects (i.e. ordinary numbers if N=2). | |
| 996 """ | |
| 997 | |
|
188
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
998 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
|
999 """ |
|
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1000 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
|
1001 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
|
1002 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
|
1003 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
|
1004 """ |
|
188
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
1005 ArrayFieldsDataSet.__init__(self, **kwargs) |
| 41 | 1006 self.data=data_array |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1007 self.fields_columns=fields_columns |
| 41 | 1008 |
| 1009 # check consistency and complete slices definitions | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1010 for fieldname, fieldcolumns in self.fields_columns.items(): |
| 41 | 1011 if type(fieldcolumns) is int: |
| 1012 assert fieldcolumns>=0 and fieldcolumns<data_array.shape[1] | |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
1013 if 1: |
|
227
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1014 #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
|
1015 # 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
|
1016 # 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
|
1017 # - James 22/05/2008 |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1018 self.fields_columns[fieldname]=[fieldcolumns] |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1019 else: |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1020 self.fields_columns[fieldname]=fieldcolumns |
| 41 | 1021 elif type(fieldcolumns) is slice: |
| 1022 start,step=None,None | |
| 1023 if not fieldcolumns.start: | |
| 1024 start=0 | |
| 1025 if not fieldcolumns.step: | |
| 1026 step=1 | |
| 1027 if start or step: | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1028 self.fields_columns[fieldname]=slice(start,fieldcolumns.stop,step) |
| 41 | 1029 elif hasattr(fieldcolumns,"__iter__"): # something like a list |
| 1030 for i in fieldcolumns: | |
| 1031 assert i>=0 and i<data_array.shape[1] | |
| 1032 | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1033 def fieldNames(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1034 return self.fields_columns.keys() |
| 41 | 1035 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1036 def __len__(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1037 return len(self.data) |
| 41 | 1038 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1039 def dontuse__getitem__(self,key): |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1040 """More efficient implementation than the default __getitem__""" |
|
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1041 fieldnames=self.fields_columns.keys() |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1042 values=self.fields_columns.values() |
| 80 | 1043 if type(key) is int: |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1044 return Example(fieldnames, |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1045 [self.data[key,col] for col in values]) |
| 80 | 1046 if type(key) is slice: |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1047 return MinibatchDataSet(Example(fieldnames, |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1048 [self.data[key,col] for col in values])) |
| 80 | 1049 if type(key) is list: |
| 1050 for i in range(len(key)): | |
| 1051 if self.hasFields(key[i]): | |
| 1052 key[i]=self.fields_columns[key[i]] | |
| 1053 return MinibatchDataSet(Example(fieldnames, | |
|
88
6749d18e11c8
bugfix as numpy numpy don't support self.data[[i1,...],[i2,...]] when their is more then two i1 and i2
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
85
diff
changeset
|
1054 #we must separate differently for list as numpy |
| 128 | 1055 # doesn't support self.data[[i1,...],[i2,...]] |
|
88
6749d18e11c8
bugfix as numpy numpy don't support self.data[[i1,...],[i2,...]] when their is more then two i1 and i2
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
85
diff
changeset
|
1056 # when their is more then two i1 and i2 |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1057 [self.data[key,:][:,col] |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1058 if isinstance(col,list) else |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1059 self.data[key,col] for col in values]), |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1060 |
| 128 | 1061 |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1062 self.valuesVStack,self.valuesHStack) |
| 80 | 1063 |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1064 # else check for a fieldname |
| 80 | 1065 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
|
1066 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
|
1067 # else we are trying to access a property of the dataset |
| 80 | 1068 assert key in self.__dict__ # else it means we are trying to access a non-existing property |
| 1069 return self.__dict__[key] | |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1070 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1071 def dontuse__iter__(self): |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1072 class ArrayDataSetIteratorIter(object): |
|
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1073 def __init__(self,dataset,fieldnames): |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1074 if fieldnames is None: fieldnames = dataset.fieldNames() |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1075 # 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
|
1076 self.minibatch = Example(fieldnames,[0]*len(fieldnames)) |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1077 self.dataset=dataset |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1078 self.current=0 |
|
238
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1079 self.columns = [self.dataset.fields_columns[f] |
|
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1080 for f in self.minibatch._names] |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1081 self.l = self.dataset.data.shape[0] |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1082 def __iter__(self): |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1083 return self |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1084 def next(self): |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1085 #@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
|
1086 # Otherwise, MinibatchWrapAroundIterator do it. |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1087 if self.current>=self.l: |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1088 raise StopIteration |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1089 sub_data = self.dataset.data[self.current] |
|
238
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1090 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
|
1091 |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1092 self.current+=1 |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1093 return self.minibatch |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1094 |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1095 return ArrayDataSetIteratorIter(self,self.fieldNames()) |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1096 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1097 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
|
1098 cursor = Example(fieldnames,[0]*len(fieldnames)) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1099 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
|
1100 for n in xrange(n_batches): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1101 if offset == len(self): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1102 break |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1103 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
|
1104 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
|
1105 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
|
1106 yield cursor |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1107 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1108 #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
|
1109 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1110 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1111 class CachedDataSet(DataSet): |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1112 """ |
| 167 | 1113 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
|
1114 (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
|
1115 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
|
1116 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
|
1117 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1118 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
|
1119 (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
|
1120 first access. |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1121 |
| 167 | 1122 @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
|
1123 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
|
1124 |
| 167 | 1125 @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
|
1126 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
|
1127 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
|
1128 """ |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1129 def __init__(self,source_dataset,cache_all_upon_construction=False): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1130 self.source_dataset=source_dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1131 self.cache_all_upon_construction=cache_all_upon_construction |
|
152
3f627e844cba
Fixes in CacheDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
151
diff
changeset
|
1132 self.cached_examples = [] |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1133 if cache_all_upon_construction: |
|
77
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1134 # this potentially brings all the source examples |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1135 # into memory at once, which may be too much |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1136 # the work could possibly be done by minibatches |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1137 # 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
|
1138 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
|
1139 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
|
1140 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
|
1141 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
|
1142 |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1143 self.fieldNames = source_dataset.fieldNames |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1144 self.hasFields = source_dataset.hasFields |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1145 self.valuesHStack = source_dataset.valuesHStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1146 self.valuesVStack = source_dataset.valuesVStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1147 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1148 def __len__(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1149 return len(self.source_dataset) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1150 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1151 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1152 class CacheIterator(object): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1153 def __init__(self,dataset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1154 self.dataset=dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1155 self.current=offset |
|
254
8ec867d12428
optimication in CachedDataSet.minibatches_nowrap
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
253
diff
changeset
|
1156 self.all_fields = self.dataset.fieldNames()==fieldnames |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1157 def __iter__(self): return self |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1158 def next(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1159 upper = self.current+minibatch_size |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1160 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
|
1161 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
|
1162 # cache everything from current length to upper |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1163 for example in self.dataset.source_dataset[cache_len:upper]: |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1164 self.dataset.cached_examples.append(example) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1165 all_fields_minibatch = Example(self.dataset.fieldNames(), |
|
152
3f627e844cba
Fixes in CacheDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
151
diff
changeset
|
1166 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
|
1167 self.current+=minibatch_size |
|
254
8ec867d12428
optimication in CachedDataSet.minibatches_nowrap
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
253
diff
changeset
|
1168 if self.all_fields: |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1169 return all_fields_minibatch |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1170 return Example(fieldnames,[all_fields_minibatch[name] for name in fieldnames]) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1171 return CacheIterator(self) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1172 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1173 def dontuse__getitem__(self,i): |
| 153 | 1174 if type(i)==int and len(self.cached_examples)>i: |
| 1175 return self.cached_examples[i] | |
| 1176 else: | |
|
251
7e6edee187e3
optimization of CachedDataSet__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
243
diff
changeset
|
1177 return self.source_dataset[i] |
|
252
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1178 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1179 def __iter__(self): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1180 class CacheIteratorIter(object): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1181 def __init__(self,dataset): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1182 self.dataset=dataset |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1183 self.l = len(dataset) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1184 self.current = 0 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1185 self.fieldnames = self.dataset.fieldNames() |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1186 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
|
1187 def __iter__(self): return self |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1188 def next(self): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1189 if self.current>=self.l: |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1190 raise StopIteration |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1191 cache_len = len(self.dataset.cached_examples) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1192 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
|
1193 # cache everything from current length to upper |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1194 self.dataset.cached_examples.append( |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1195 self.dataset.source_dataset[self.current]) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1196 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
|
1197 self.current+=1 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1198 return self.example |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1199 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1200 return CacheIteratorIter(self) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1201 |
|
57
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1202 class ApplyFunctionDataSet(DataSet): |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1203 """ |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1204 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
|
1205 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
|
1206 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
|
1207 (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
|
1208 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1209 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
|
1210 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1211 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
|
1212 (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
|
1213 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
|
1214 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
|
1215 (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
|
1216 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
|
1217 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
|
1218 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1219 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
|
1220 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
|
1221 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1222 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
|
1223 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
|
1224 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1225 """ |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1226 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1227 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
|
1228 values_hstack=None,values_vstack=None, |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1229 description=None,fieldtypes=None): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1230 """ |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1231 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
|
1232 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
|
1233 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
|
1234 (provided in a list). |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1235 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1236 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
|
1237 (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
|
1238 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
|
1239 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1240 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
|
1241 (for the new dataset created)? |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1242 """ |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1243 self.input_dataset=input_dataset |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1244 self.function=function |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1245 self.output_names=output_names |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1246 self.minibatch_mode=minibatch_mode |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1247 DataSet.__init__(self,description,fieldtypes) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1248 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
|
1249 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
|
1250 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1251 def __len__(self): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1252 return len(self.input_dataset) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1253 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1254 def fieldNames(self): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1255 return self.output_names |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1256 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1257 def minibatches_nowrap(self, fieldnames, *args, **kwargs): |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1258 all_input_fieldNames = self.input_dataset.fieldNames() |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1259 mbnw = self.input_dataset.minibatches_nowrap |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1260 |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1261 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
|
1262 if self.minibatch_mode: |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1263 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
|
1264 else: |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1265 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
|
1266 output_examples = [self.function(*input_example) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1267 for input_example in input_examples] |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1268 all_output_fields = zip(*output_examples) |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1269 |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1270 all_outputs = Example(self.output_names, all_output_fields) |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1271 #print 'input_fields', input_fields |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1272 #print 'all_outputs', all_outputs |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1273 if fieldnames==self.output_names: |
| 293 | 1274 rval = all_outputs |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1275 else: |
| 293 | 1276 rval = Example(fieldnames,[all_outputs[name] for name in fieldnames]) |
|
296
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1277 #print 'rval', rval |
|
f5d33f9c0b9c
ApplyFunctionDataSet passing
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
293
diff
changeset
|
1278 #print '--------' |
| 293 | 1279 yield rval |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1280 |
|
290
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1281 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
|
1282 class ApplyFunctionSingleExampleIterator(object): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1283 def __init__(self,output_dataset): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1284 self.current=0 |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1285 self.output_dataset=output_dataset |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1286 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
|
1287 def __iter__(self): return self |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1288 def next(self): |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1289 if self.output_dataset.minibatch_mode: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1290 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
|
1291 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
|
1292 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
|
1293 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
|
1294 else: |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1295 function_inputs = self.input_iterator.next() |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1296 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
|
1297 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
|
1298 return ApplyFunctionSingleExampleIterator(self) |
|
9b533cc7874a
trying to get default implemenations to work
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
274
diff
changeset
|
1299 |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1300 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
|
1301 """ |
| 167 | 1302 Wraps an arbitrary L{DataSet} into one for supervised learning tasks |
| 1303 by forcing the user to define a set of fields as the 'input' field | |
| 1304 and a set of fields as the 'target' field. Optionally, a single | |
| 1305 weight_field can also be defined. | |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1306 """ |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1307 args = ((input_fields,'input'),(output_fields,'target')) |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1308 if weight_field: args+=(([weight_field],'weight')) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
1309 return src_dataset.merge_fields(*args) |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1310 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1311 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1312 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1313 |