Mercurial > pylearn
annotate dataset.py @ 282:b55c829695f1
renamed test_filetensor.py to _test_filetensor.py to have it being executed by autotest
| author | Frederic Bastien <bastienf@iro.umontreal.ca> |
|---|---|
| date | Fri, 06 Jun 2008 13:46:10 -0400 |
| parents | ed70580f2324 |
| children | 9b533cc7874a |
| 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 |
|
12
ff4e551490f1
Added LookupList type in lookup_list.py and used it to keep order
bengioy@esprit.iro.umontreal.ca
parents:
11
diff
changeset
|
2 from lookup_list import LookupList |
|
ff4e551490f1
Added LookupList type in lookup_list.py and used it to keep order
bengioy@esprit.iro.umontreal.ca
parents:
11
diff
changeset
|
3 Example = LookupList |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
4 from misc import unique_elements_list_intersection |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
5 from string import join |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
6 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
|
7 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
|
8 |
| 166 | 9 from exceptions import * |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
10 |
|
110
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
11 class AttributesHolder(object): |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
12 def __init__(self): pass |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
13 |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
14 def attributeNames(self): |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
15 raise AbstractFunction() |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
16 |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
17 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
|
18 """ |
|
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
19 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
|
20 """ |
|
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
21 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
|
22 attribute_values = [attribute_values] |
|
110
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
23 if make_copies: |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
24 for name,value in zip(attribute_names,attribute_values): |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
25 self.__setattr__(name,copy.deepcopy(value)) |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
26 else: |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
27 for name,value in zip(attribute_names,attribute_values): |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
28 self.__setattr__(name,value) |
|
193
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
29 |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
30 def getAttributes(self,attribute_names=None, return_copy=False): |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
31 """ |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
32 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
|
33 """ |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
34 if attribute_names is None: |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
35 attribute_names = self.attributeNames() |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
36 if return_copy: |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
37 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
|
38 else: |
|
cb6b945acf5a
Complete redesign of learner...
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
188
diff
changeset
|
39 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
|
40 |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
41 |
|
8fa1ef2411a0
Worked on OneShotTLearner and implementation of LinearRegression
bengioy@bengiomac.local
parents:
105
diff
changeset
|
42 class DataSet(AttributesHolder): |
| 16 | 43 """A virtual base class for datasets. |
| 44 | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
45 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
|
46 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
|
47 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
|
48 python object, which depends on the particular dataset. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
49 |
|
241
ddb88a8e9fd2
If I understand properly, the length of an unbounded stream is sys.maxint
delallea@opale.iro.umontreal.ca
parents:
231
diff
changeset
|
50 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
|
51 should return sys.maxint). |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
52 |
| 16 | 53 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
|
54 examples or the fields in a variety of ways. A DataSet need not necessarily have a finite |
| 16 | 55 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
|
56 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
|
57 feasible or not recommended on streams). |
| 16 | 58 |
| 59 To iterate over examples, there are several possibilities: | |
|
90
a289b8bed64c
corrected comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
88
diff
changeset
|
60 - for example in dataset: |
|
a289b8bed64c
corrected comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
88
diff
changeset
|
61 - for val1,val2,... in dataset: |
|
a289b8bed64c
corrected comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
88
diff
changeset
|
62 - for example in dataset(field1, field2,field3, ...): |
|
a289b8bed64c
corrected comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
88
diff
changeset
|
63 - for val1,val2,val3 in dataset(field1, field2,field3): |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
64 - for minibatch in dataset.minibatches([field1, field2, ...],minibatch_size=N): |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
65 - 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
|
66 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
|
67 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
|
68 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
|
69 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
|
70 next minibatch returned). This is important because these iterators |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
71 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
|
72 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
|
73 dataset length. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
74 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
75 To iterate over fields, one can do |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
76 - for field in dataset.fields(): |
|
46
c5b07e87b0cb
comments modif made by Yoshua
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
45
diff
changeset
|
77 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
|
78 - 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
|
79 - 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
|
80 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
|
81 - for field_examples in dataset.fields(): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
82 for example_value in field_examples: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
83 ... |
|
241
ddb88a8e9fd2
If I understand properly, the length of an unbounded stream is sys.maxint
delallea@opale.iro.umontreal.ca
parents:
231
diff
changeset
|
84 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
|
85 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
|
86 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
|
87 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
|
88 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
|
89 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
|
90 dataset2 = dataset1.fields().examples() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
91 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
|
92 |
| 16 | 93 Note: Fields are not mutually exclusive, i.e. two fields can overlap in their actual content. |
| 94 | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
95 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
|
96 (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
|
97 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
|
98 What about non-numeric values? None. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
99 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
100 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
|
101 of examples) can be extracted. These operations are not supported |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
102 by default in the case of streams. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
103 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
104 - dataset[:n] returns a dataset with the n first examples. |
| 16 | 105 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
106 - 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
|
107 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
108 - dataset[i] returns an Example. |
|
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[[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
|
111 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
112 - dataset.<property> returns the value of a property associated with |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
113 the name <property>. The following properties should be supported: |
| 41 | 114 - '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
|
115 - 'fieldtypes': a list of types (one per field) |
| 78 | 116 A DataSet may have other attributes that it makes visible to other objects. These are |
| 117 used to store information that is not example-wise but global to the dataset. | |
| 118 The list of names of these attributes is given by the attribute_names() method. | |
| 41 | 119 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
120 Datasets can be concatenated either vertically (increasing the length) or |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
121 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
|
122 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
|
123 and numpy.vstack): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
124 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
125 - 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
|
126 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
127 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
|
128 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
|
129 |
|
72
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
130 - dataset1 & dataset2 & dataset3 == dataset.vstack([dataset1,dataset2,dataset3]) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
131 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
132 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
|
133 (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
|
134 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
|
135 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
136 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
|
137 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
|
138 a DataSetFields fields1 and fields2, and fields1 | fields2 concatenates their |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
139 examples. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
140 |
| 41 | 141 A dataset can hold arbitrary key-value pairs that may be used to access meta-data |
| 142 or other properties of the dataset or associated with the dataset or the result | |
| 143 of a computation stored in a dataset. These can be accessed through the [key] syntax | |
| 144 when key is a string (or more specifically, neither an integer, a slice, nor a list). | |
| 78 | 145 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
146 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
|
147 - __len__ if it is not a stream |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
148 - fieldNames |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
149 - minibatches_nowrap (called by DataSet.minibatches()) |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
150 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
|
151 - valuesHStack |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
152 - valuesVStack |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
153 - hasFields |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
154 - __getitem__ may not be feasible with some streams |
|
2b6656b2ef52
Changed docs slightly
Joseph Turian <turian@iro.umontreal.ca>
parents:
66
diff
changeset
|
155 - __iter__ |
| 78 | 156 A sub-class should also append attributes to self._attribute_names |
| 157 (the default value returned by attributeNames()). | |
| 158 By convention, attributes not in attributeNames() should have a name | |
| 159 starting with an underscore. | |
| 160 @todo enforce/test that convention! | |
|
2
3fddb1c8f955
Rewrote DataSet interface and created FiniteDataSet interface.
bengioy@bengiomac.local
parents:
1
diff
changeset
|
161 """ |
|
1
2cd82666b9a7
Added statscollector and started writing dataset and learner.
bengioy@esprit.iro.umontreal.ca
parents:
0
diff
changeset
|
162 |
| 83 | 163 numpy_vstack = lambda fieldname,values: numpy.vstack(values) |
| 164 numpy_hstack = lambda fieldnames,values: numpy.hstack(values) | |
|
77
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
165 |
|
57
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
166 def __init__(self,description=None,fieldtypes=None): |
| 41 | 167 if description is None: |
| 168 # by default return "<DataSetType>(<SuperClass1>,<SuperClass2>,...)" | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
169 description = type(self).__name__ + " ( " + join([x.__name__ for x in type(self).__bases__]) + " )" |
| 41 | 170 self.description=description |
| 60 | 171 self.fieldtypes=fieldtypes |
| 78 | 172 self._attribute_names = ["description"] |
| 173 if fieldtypes: | |
| 174 self._attribute_names.append("fieldtypes") | |
| 175 | |
| 176 def attributeNames(self): return self._attribute_names | |
| 177 | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
178 class MinibatchToSingleExampleIterator(object): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
179 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
180 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
|
181 single-example values in the result. Therefore the result of |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
182 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
|
183 (whereas the result of iterating over minibatches gives in each |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
184 Example field an iterable object over the individual examples in |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
185 the minibatch). |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
186 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
187 def __init__(self, minibatch_iterator): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
188 self.minibatch_iterator = minibatch_iterator |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
189 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
|
190 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
|
191 return self |
|
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
192 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
|
193 size1_minibatch = self.minibatch_iterator.next() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
194 if not self.minibatch: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
195 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
|
196 else: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
197 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
|
198 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
|
199 |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
200 def next_index(self): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
201 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
|
202 |
|
3
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
203 def __iter__(self): |
| 16 | 204 """Supports the syntax "for i in dataset: ..." |
|
1
2cd82666b9a7
Added statscollector and started writing dataset and learner.
bengioy@esprit.iro.umontreal.ca
parents:
0
diff
changeset
|
205 |
| 16 | 206 Using this syntax, "i" will be an Example instance (or equivalent) with |
| 207 all the fields of DataSet self. Every field of "i" will give access to | |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
208 a field of a single example. Fields should be accessible via |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
209 i["fielname"] or i[3] (in the order defined by the elements of the |
|
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
210 Example returned by this iterator), but the derived class is free |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
211 to accept any type of identifier, and add extra functionality to the iterator. |
| 16 | 212 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
213 The default implementation calls the minibatches iterator and extracts the first example of each field. |
|
11
be128b9127c8
Debugged (to the extent of my tests) the new version of dataset
bengioy@esprit.iro.umontreal.ca
parents:
9
diff
changeset
|
214 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
215 return DataSet.MinibatchToSingleExampleIterator(self.minibatches(None, minibatch_size = 1)) |
|
2
3fddb1c8f955
Rewrote DataSet interface and created FiniteDataSet interface.
bengioy@bengiomac.local
parents:
1
diff
changeset
|
216 |
|
188
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
217 def __contains__(self, fieldname): |
|
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
218 return (fieldname in self.fieldNames()) \ |
|
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
219 or (fieldname in self.attributeNames()) |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
220 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
221 class MinibatchWrapAroundIterator(object): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
222 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
223 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
|
224 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
|
225 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
|
226 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
|
227 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
|
228 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
|
229 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
230 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
|
231 self.dataset=dataset |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
232 self.fieldnames=fieldnames |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
233 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
|
234 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
|
235 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
|
236 self.next_row=offset |
|
98
7186e4f502d1
bugfix in DataSet.minibatch to correctly wrap in all corner case.
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
95
diff
changeset
|
237 self.offset=offset |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
238 self.L=len(dataset) |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
239 assert offset+minibatch_size<=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
|
240 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
|
241 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
|
242 ds_nbatches = min(n_batches,ds_nbatches) |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
243 if fieldnames: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
244 assert dataset.hasFields(*fieldnames) |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
245 else: |
|
98
7186e4f502d1
bugfix in DataSet.minibatch to correctly wrap in all corner case.
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
95
diff
changeset
|
246 self.fieldnames=dataset.fieldNames() |
|
7186e4f502d1
bugfix in DataSet.minibatch to correctly wrap in all corner case.
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
95
diff
changeset
|
247 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
|
248 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
|
249 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
250 def __iter__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
251 return self |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
252 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
253 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
|
254 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
|
255 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
256 def next(self): |
|
43
e92244f30116
Corrected iterator logic errors
bengioy@grenat.iro.umontreal.ca
parents:
42
diff
changeset
|
257 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
|
258 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
|
259 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
|
260 raise StopIteration |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
261 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
|
262 if upper <=self.L: |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
263 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
|
264 else: |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
265 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
|
266 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
|
267 # 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
|
268 # 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
|
269 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
|
270 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
|
271 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
|
272 # 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
|
273 # 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
|
274 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
|
275 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
|
276 minibatch = Example(self.fieldnames, |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
277 [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
|
278 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
|
279 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
|
280 self.n_batches_done+=1 |
|
43
e92244f30116
Corrected iterator logic errors
bengioy@grenat.iro.umontreal.ca
parents:
42
diff
changeset
|
281 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
|
282 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
|
283 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
|
284 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
|
285 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
|
286 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
|
287 ds_nbatches,self.next_row) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
288 return DataSetFields(MinibatchDataSet(minibatch,self.dataset.valuesVStack, |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
289 self.dataset.valuesHStack), |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
290 minibatch.keys()) |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
291 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
292 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
293 minibatches_fieldnames = None |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
294 minibatches_minibatch_size = 1 |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
295 minibatches_n_batches = None |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
296 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
|
297 fieldnames = minibatches_fieldnames, |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
298 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
|
299 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
|
300 offset = 0): |
|
6
d5738b79089a
Removed MinibatchIterator and instead made minibatch_size a field of all DataSets,
bengioy@bengiomac.local
parents:
5
diff
changeset
|
301 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
302 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
|
303 |
|
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
304 for i in dataset.minibatches(None,**kwargs): ... |
| 16 | 305 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
306 for i in dataset.minibatches([f1, f2, f3],**kwargs): ... |
| 16 | 307 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
308 for i1, i2, i3 in dataset.minibatches([f1, f2, f3],**kwargs): ... |
| 16 | 309 |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
310 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
|
311 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
|
312 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
|
313 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
|
314 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
|
315 |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
316 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
|
317 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
|
318 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
|
319 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
320 The minibatches iterator is expected to return upon each call to next() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
321 a DataSetFields object, which is a LookupList (indexed by the field names) whose |
| 80 | 322 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
|
323 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
|
324 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
|
325 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
|
326 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
327 PARAMETERS |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
328 - 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
|
329 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
|
330 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
|
331 derived class can choose a default, e.g. all fields. |
| 16 | 332 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
333 - minibatch_size (integer, default 1) |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
334 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
|
335 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
|
336 |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
337 - n_batches (integer, default None) |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
338 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
|
339 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
|
340 iterator should support looping indefinitely. |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
341 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
342 - 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
|
343 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
|
344 |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
345 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
|
346 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
|
347 |
|
11
be128b9127c8
Debugged (to the extent of my tests) the new version of dataset
bengioy@esprit.iro.umontreal.ca
parents:
9
diff
changeset
|
348 """ |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
349 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
|
350 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
351 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
|
352 """ |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
353 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
|
354 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
|
355 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
|
356 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
|
357 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
|
358 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
|
359 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
|
360 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
|
361 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
|
362 """ |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
363 raise AbstractFunction() |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
364 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
365 def __len__(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
366 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
367 len(dataset) returns the number of examples in the dataset. |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
368 By default, a DataSet is a 'stream', i.e. it has an unbounded length (sys.maxint). |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
369 Sub-classes which implement finite-length datasets should redefine this method. |
|
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
|
370 Some methods only make sense for finite-length datasets. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
371 """ |
| 123 | 372 return maxint |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
373 |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
374 def is_unbounded(self): |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
375 """ |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
376 Tests whether a dataset is unbounded (e.g. a stream). |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
377 """ |
| 123 | 378 return len(self)==maxint |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
379 |
|
26
672fe4b23032
Fixed dataset errors so that _test_dataset.py works again.
bengioy@grenat.iro.umontreal.ca
parents:
23
diff
changeset
|
380 def hasFields(self,*fieldnames): |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
381 """ |
|
22
b6b36f65664f
Created virtual sub-classes of DataSet: {Finite{Length,Width},Sliceable}DataSet,
bengioy@esprit.iro.umontreal.ca
parents:
20
diff
changeset
|
382 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
|
383 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
|
384 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
|
385 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
386 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
|
387 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
|
388 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
389 return len(unique_elements_list_intersection(fieldnames,self.fieldNames()))>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
390 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
391 def fieldNames(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
392 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
393 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
|
394 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
|
395 """ |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
396 raise AbstractFunction() |
|
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
397 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
398 def __call__(self,*fieldnames): |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
399 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
400 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
|
401 """ |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
402 assert self.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
403 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
|
404 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
405 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
|
406 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
407 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
|
408 """ |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
409 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
|
410 |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
411 def getitem_key(self, fieldname): |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
412 """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
|
413 getitem. |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
414 """ |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
415 #removing as per discussion June 4. --JSB |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
416 |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
417 i = fieldname |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
418 # else check for a fieldname |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
419 if self.hasFields(i): |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
420 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
|
421 # 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
|
422 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
|
423 return self.__dict__[i] |
|
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
424 |
|
2
3fddb1c8f955
Rewrote DataSet interface and created FiniteDataSet interface.
bengioy@bengiomac.local
parents:
1
diff
changeset
|
425 def __getitem__(self,i): |
|
28
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
426 """ |
|
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
427 dataset[i] returns the (i+1)-th example of the dataset. |
|
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
428 dataset[i:j] returns the subdataset with examples i,i+1,...,j-1. |
|
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
429 dataset[i:j:s] returns the subdataset with examples i,i+2,i+4...,j-2. |
|
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
430 dataset[[i1,i2,..,in]] returns the subdataset with examples i1,i2,...,in. |
| 41 | 431 dataset['key'] returns a property associated with the given 'key' string. |
| 432 If 'key' is a fieldname, then the VStacked field values (iterable over | |
| 433 field values) for that field is returned. Other keys may be supported | |
| 434 by different dataset subclasses. The following key names are encouraged: | |
| 435 - 'description': a textual description or name for the dataset | |
| 436 - '<fieldname>.type': a type name or value for a given <fieldname> | |
|
1
2cd82666b9a7
Added statscollector and started writing dataset and learner.
bengioy@esprit.iro.umontreal.ca
parents:
0
diff
changeset
|
437 |
| 39 | 438 Note that some stream datasets may be unable to implement random access, i.e. |
| 439 arbitrary slicing/indexing | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
440 because they can only iterate through examples one or a minibatch at a time |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
441 and do not actually store or keep 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
|
442 |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
443 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
|
444 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
|
445 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
|
446 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
|
447 """ |
| 41 | 448 # check for an index |
|
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 if type(i) is int: |
|
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 return DataSet.MinibatchToSingleExampleIterator( |
|
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 self.minibatches(minibatch_size=1,n_batches=1,offset=i)).next() |
| 41 | 452 rows=None |
| 453 # or a slice | |
|
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
|
454 if type(i) is slice: |
|
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
|
455 #print 'i=',i |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
456 if not i.start: i=slice(0,i.stop,i.step) |
|
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
|
457 if not i.stop: i=slice(i.start,len(self),i.step) |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
458 if not i.step: i=slice(i.start,i.stop,1) |
|
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
|
459 if i.step is 1: |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
460 return self.minibatches(minibatch_size=i.stop-i.start,n_batches=1,offset=i.start).next().examples() |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
461 rows = range(i.start,i.stop,i.step) |
| 41 | 462 # or a list of indices |
| 463 elif type(i) is list: | |
|
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
|
464 rows = i |
| 41 | 465 if rows is not None: |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
466 examples = [self[row] for row in rows] |
|
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
467 fields_values = zip(*examples) |
|
45
a5c70dc42972
Test functions for dataset.py
bengioy@grenat.iro.umontreal.ca
parents:
44
diff
changeset
|
468 return MinibatchDataSet( |
| 41 | 469 Example(self.fieldNames(),[ self.valuesVStack(fieldname,field_values) |
| 470 for fieldname,field_values | |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
471 in zip(self.fieldNames(),fields_values)]), |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
472 self.valuesVStack,self.valuesHStack) |
|
269
fdce496c3b56
deprecating __getitem__[fieldname] syntax
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
268
diff
changeset
|
473 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
|
474 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
475 def valuesHStack(self,fieldnames,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
476 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
477 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
|
478 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
|
479 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
|
480 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
|
481 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
|
482 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
|
483 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
|
484 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
|
485 pointing to the original values for other data types. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
486 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
487 all_numpy=True |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
488 for value in fieldvalues: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
489 if not type(value) is numpy.ndarray: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
490 all_numpy=False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
491 if all_numpy: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
492 return numpy.hstack(fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
493 # 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
|
494 return fieldvalues |
|
26
672fe4b23032
Fixed dataset errors so that _test_dataset.py works again.
bengioy@grenat.iro.umontreal.ca
parents:
23
diff
changeset
|
495 |
|
672fe4b23032
Fixed dataset errors so that _test_dataset.py works again.
bengioy@grenat.iro.umontreal.ca
parents:
23
diff
changeset
|
496 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
497 def valuesVStack(self,fieldname,values): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
498 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
499 Return a value that corresponds to concatenating (vertically) several values of the |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
500 same field. This can be important to build a minibatch out of individual examples. This |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
501 is likely 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
|
502 result should be numpy.vstack(values). |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
503 The default is to use numpy.vstack for numpy.ndarray values, and a list |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
504 pointing to the original values for other data types. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
505 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
506 all_numpy=True |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
507 for value in values: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
508 if not type(value) is numpy.ndarray: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
509 all_numpy=False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
510 if all_numpy: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
511 return numpy.vstack(values) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
512 # the default implementation of vertical stacking is to put values in a list |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
513 return values |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
514 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
515 def __or__(self,other): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
516 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
517 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
|
518 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
|
519 """ |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
520 return HStackedDataSet([self,other]) |
|
3
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
521 |
|
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
|
522 def __and__(self,other): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
523 """ |
|
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
|
524 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
|
525 (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
|
526 works if they all have the same fields. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
527 """ |
|
135
0d8e721cc63c
Fixed bugs in dataset to make test_mlp.py work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
134
diff
changeset
|
528 return VStackedDataSet([self,other]) |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
529 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
530 def hstack(datasets): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
531 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
532 hstack(dataset1,dataset2,...) returns dataset1 | datataset2 | ... |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
533 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
|
534 of the individual datasets. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
535 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
536 assert len(datasets)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
537 if len(datasets)==1: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
538 return datasets[0] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
539 return HStackedDataSet(datasets) |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
540 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
541 def vstack(datasets): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
542 """ |
|
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
|
543 vstack(dataset1,dataset2,...) returns dataset1 & datataset2 & ... |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
544 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
|
545 over those of dataset2, etc. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
546 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
547 assert len(datasets)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
548 if len(datasets)==1: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
549 return datasets[0] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
550 return VStackedDataSet(datasets) |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
551 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
552 class FieldsSubsetDataSet(DataSet): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
553 """ |
| 167 | 554 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
|
555 """ |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
556 def __init__(self,src,fieldnames): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
557 self.src=src |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
558 self.fieldnames=fieldnames |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
559 assert src.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
560 self.valuesHStack = src.valuesHStack |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
561 self.valuesVStack = src.valuesVStack |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
562 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
563 def __len__(self): return len(self.src) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
564 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
565 def fieldNames(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
566 return self.fieldnames |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
567 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
568 def __iter__(self): |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
569 class FieldsSubsetIterator(object): |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
570 def __init__(self,ds): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
571 self.ds=ds |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
572 self.src_iter=ds.src.__iter__() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
573 self.example=None |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
574 def __iter__(self): return self |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
575 def next(self): |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
576 complete_example = self.src_iter.next() |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
577 if self.example: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
578 self.example._values=[complete_example[field] |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
579 for field in self.ds.fieldnames] |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
580 else: |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
581 self.example=Example(self.ds.fieldnames, |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
582 [complete_example[field] for field in self.ds.fieldnames]) |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
583 return self.example |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
584 return FieldsSubsetIterator(self) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
585 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
586 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
|
587 assert self.hasFields(*fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
588 return self.src.minibatches_nowrap(fieldnames,minibatch_size,n_batches,offset) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
589 def __getitem__(self,i): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
590 return FieldsSubsetDataSet(self.src[i],self.fieldnames) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
591 |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
592 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
593 class DataSetFields(LookupList): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
594 """ |
| 167 | 595 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
|
596 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
|
597 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
|
598 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
599 To iterate over fields, one can do |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
600 * for fields in dataset.fields() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
601 * 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
|
602 * 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
|
603 and each of these fields is iterable over the examples: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
604 * for field_examples in dataset.fields(): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
605 for example_value in field_examples: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
606 ... |
|
241
ddb88a8e9fd2
If I understand properly, the length of an unbounded stream is sys.maxint
delallea@opale.iro.umontreal.ca
parents:
231
diff
changeset
|
607 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
|
608 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
|
609 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
|
610 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
|
611 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
|
612 be turned back into a DataSet with its examples() method: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
613 dataset2 = dataset1.fields().examples() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
614 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
|
615 |
|
88fd1cce08b9
replaced infinity for length by raise UnboundedDataSet and use & instead of + to concatenate datasets
bengioy@esprit.iro.umontreal.ca
parents:
39
diff
changeset
|
616 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
|
617 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
|
618 the examples. |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
619 """ |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
620 def __init__(self,dataset,fieldnames): |
|
65
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
621 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
|
622 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
|
623 fieldnames=dataset.fieldNames() |
|
274
ed70580f2324
bugfix in FieldSubsetDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
273
diff
changeset
|
624 elif not list(fieldnames)==list(dataset.fieldNames()): |
|
ed70580f2324
bugfix in FieldSubsetDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
273
diff
changeset
|
625 #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
|
626 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
|
627 assert dataset.hasFields(*fieldnames) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
628 self.dataset=dataset |
|
66
dde1fb1b63ba
fixed test and removed print
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
65
diff
changeset
|
629 |
|
64
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
630 if isinstance(dataset,MinibatchDataSet): |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
631 LookupList.__init__(self,fieldnames,list(dataset._fields)) |
|
65
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
632 elif isinstance(original_dataset,MinibatchDataSet): |
|
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
633 LookupList.__init__(self,fieldnames, |
|
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
634 [original_dataset._fields[field] |
|
d48eba49a2f4
fixed the infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
64
diff
changeset
|
635 for field in fieldnames]) |
|
64
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
636 else: |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
637 minibatch_iterator = dataset.minibatches(fieldnames, |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
638 minibatch_size=len(dataset), |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
639 n_batches=1) |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
640 minibatch=minibatch_iterator.next() |
|
863da25a60f1
trying to fix infinite loop
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
62
diff
changeset
|
641 LookupList.__init__(self,fieldnames,minibatch) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
642 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
643 def examples(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
644 return self.dataset |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
645 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
646 def __or__(self,other): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
647 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
648 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
|
649 of the list of examples of DataSetFields fields1 and fields2. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
650 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
651 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
|
652 |
|
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
|
653 def __and__(self,other): |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
654 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
655 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
|
656 of the fields of DataSetFields fields1 and fields2. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
657 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
658 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
|
659 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
660 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
661 class MinibatchDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
662 """ |
| 167 | 663 Turn a L{LookupList} 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
|
664 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
|
665 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
666 def __init__(self,fields_lookuplist,values_vstack=DataSet().valuesVStack, |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
667 values_hstack=DataSet().valuesHStack): |
|
17
759d17112b23
more comments, looping ArrayDataSet iterator, bugfixes to lookup_list, more tests
bergstrj@iro.umontreal.ca
diff
changeset
|
668 """ |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
669 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
|
670 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
|
671 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
|
672 """ |
|
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
|
673 |
|
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
|
674 self._fields=fields_lookuplist |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
675 assert len(fields_lookuplist)>0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
676 self.length=len(fields_lookuplist[0]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
677 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
|
678 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
|
679 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
|
680 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
|
681 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
|
682 print 'field=',field |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
683 assert self.length==len(field) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
684 self.values_vstack=values_vstack |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
685 self.values_hstack=values_hstack |
|
3
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
686 |
|
378b68d5c4ad
Added first (untested) version of ArrayDataSet
bengioy@bengiomac.local
parents:
2
diff
changeset
|
687 def __len__(self): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
688 return self.length |
|
28
541a273bc89f
Removed __array__ method from dataset, whose
bengioy@grenat.iro.umontreal.ca
parents:
26
diff
changeset
|
689 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
690 def __getitem__(self,i): |
| 80 | 691 if type(i) in (slice,list): |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
692 return DataSetFields(MinibatchDataSet( |
| 80 | 693 Example(self._fields.keys(),[field[i] for field in self._fields])),self.fieldNames()) |
| 694 if type(i) is int: | |
| 85 | 695 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
|
696 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
|
697 return self._fields[i] |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
698 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
|
699 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
|
700 |
|
29
46c5c90019c2
Changed apply_function so that it propagates methods of the source.
bengioy@grenat.iro.umontreal.ca
parents:
28
diff
changeset
|
701 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
|
702 return self._fields.keys() |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
703 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
704 def hasFields(self,*fieldnames): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
705 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
|
706 if fieldname not in self._fields.keys(): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
707 return False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
708 return True |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
709 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
710 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
|
711 #@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
|
712 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
|
713 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
|
714 # 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
|
715 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
|
716 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
|
717 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
718 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
|
719 self.next_example=offset |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
720 assert minibatch_size > 0 |
| 41 | 721 if offset+minibatch_size > ds.length: |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
722 raise NotImplementedError() |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
723 def __iter__(self): |
|
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
724 return self |
|
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
725 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
|
726 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
|
727 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
|
728 #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
|
729 # [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
|
730 # 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
|
731 # 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
|
732 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
|
733 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
|
734 #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
|
735 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
|
736 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
|
737 #print minibatch |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
738 self.next_example+=minibatch_size |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
739 return minibatch |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
740 |
|
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
|
741 # 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
|
742 return Iterator(self,fieldnames) |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
743 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
744 def valuesVStack(self,fieldname,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
745 return self.values_vstack(fieldname,fieldvalues) |
|
20
266c68cb6136
Minor editions, plus adding untested ApplyFunctionDataset for GradientLearner in the works.
bengioy@bengiomac.local
parents:
19
diff
changeset
|
746 |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
747 def valuesHStack(self,fieldnames,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
748 return self.values_hstack(fieldnames,fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
749 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
750 class HStackedDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
751 """ |
| 167 | 752 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
|
753 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
|
754 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
755 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
|
756 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
|
757 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
|
758 |
| 167 | 759 @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
|
760 """ |
| 41 | 761 def __init__(self,datasets,accept_nonunique_names=False,description=None,field_types=None): |
| 762 DataSet.__init__(self,description,field_types) | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
763 self.datasets=datasets |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
764 self.accept_nonunique_names=accept_nonunique_names |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
765 self.fieldname2dataset={} |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
766 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
767 def rename_field(fieldname,dataset,i): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
768 if hasattr(dataset,"__name__"): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
769 return dataset.__name__ + "." + fieldname |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
770 return fieldname+"."+str(i) |
|
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 # 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
|
773 self.length=None |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
774 names_to_change=[] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
775 for i in xrange(len(datasets)): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
776 dataset = datasets[i] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
777 length=len(dataset) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
778 if self.length: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
779 assert self.length==length |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
780 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
781 self.length=length |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
782 for fieldname in dataset.fieldNames(): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
783 if fieldname in self.fieldname2dataset: # name conflict! |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
784 if accept_nonunique_names: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
785 fieldname=rename_field(fieldname,dataset,i) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
786 names2change.append((fieldname,i)) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
787 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
788 raise ValueError("Incompatible datasets: non-unique field name = "+fieldname) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
789 self.fieldname2dataset[fieldname]=i |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
790 for fieldname,i in names_to_change: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
791 del self.fieldname2dataset[fieldname] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
792 self.fieldname2dataset[rename_field(fieldname,self.datasets[i],i)]=i |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
793 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
794 def hasFields(self,*fieldnames): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
795 for fieldname in fieldnames: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
796 if not fieldname in self.fieldname2dataset: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
797 return False |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
798 return True |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
799 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
800 def fieldNames(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
801 return self.fieldname2dataset.keys() |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
802 |
| 41 | 803 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
|
804 |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
805 class HStackedIterator(object): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
806 def __init__(self,hsds,iterators): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
807 self.hsds=hsds |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
808 self.iterators=iterators |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
809 def __iter__(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
810 return self |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
811 def next(self): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
812 # concatenate all the fields of the minibatches |
| 140 | 813 l=LookupList() |
| 814 for iter in self.iterators: | |
| 815 l.append_lookuplist(iter.next()) | |
| 816 return l | |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
817 |
| 125 | 818 assert self.hasFields(*fieldnames) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
819 # 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
|
820 # and construct corresponding minibatch iterators |
| 140 | 821 if fieldnames and fieldnames!=self.fieldNames(): |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
822 datasets=set([]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
823 fields_in_dataset=dict([(dataset,[]) for dataset in datasets]) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
824 for fieldname in fieldnames: |
| 136 | 825 dataset=self.datasets[self.fieldname2dataset[fieldname]] |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
826 datasets.add(dataset) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
827 fields_in_dataset[dataset].append(fieldname) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
828 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
|
829 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
|
830 for dataset in datasets] |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
831 else: |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
832 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
|
833 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
|
834 return HStackedIterator(self,iterators) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
835 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
836 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
837 def valuesVStack(self,fieldname,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
838 return self.datasets[self.fieldname2dataset[fieldname]].valuesVStack(fieldname,fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
839 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
840 def valuesHStack(self,fieldnames,fieldvalues): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
841 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
842 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
|
843 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
|
844 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
|
845 all the fieldnames are of the same sub-dataset. |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
846 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
847 return self.datasets[self.fieldname2dataset[fieldnames[0]]].valuesHStack(fieldnames,fieldvalues) |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
848 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
849 class VStackedDataSet(DataSet): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
850 """ |
| 167 | 851 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
|
852 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
|
853 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
|
854 |
| 167 | 855 @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
|
856 """ |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
857 def __init__(self,datasets): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
858 self.datasets=datasets |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
859 self.length=0 |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
860 self.index2dataset={} |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
861 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
|
862 fieldnames = datasets[-1].fieldNames() |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
863 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
|
864 # 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
|
865 # 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
|
866 for dataset,k in enumerate(datasets[0:-1]): |
|
48
b6730f9a336d
Fixing MinibatchDataSet getitem
bengioy@grenat.iro.umontreal.ca
parents:
46
diff
changeset
|
867 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
|
868 L=len(dataset) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
869 for i in xrange(L): |
|
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
870 self.index2dataset[self.length+i]=k |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
871 self.datasets_start_row.append(self.length) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
872 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
|
873 assert dataset.fieldNames()==fieldnames |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
874 self.datasets_start_row.append(self.length) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
875 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
|
876 # 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
|
877 # 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
|
878 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
|
879 # 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
|
880 # TODO |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
881 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
|
882 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
883 def __len__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
884 return self.length |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
885 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
886 def fieldNames(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
887 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
|
888 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
889 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
|
890 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
|
891 |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
892 def locate_row(self,row): |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
893 """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
|
894 dataset_index = self.index2dataset[row] |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
895 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
|
896 return dataset_index, row_within_dataset |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
897 |
| 41 | 898 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
|
899 |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
900 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
|
901 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
|
902 self.vsds=vsds |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
903 self.next_row=offset |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
904 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
|
905 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
|
906 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
|
907 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
908 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
|
909 L=len(dataset) |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
910 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
|
911 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
|
912 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
|
913 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
|
914 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
|
915 n_left_in_mb=minibatch_size-L |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
916 ds_nbatches=1 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
917 else: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
918 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
|
919 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
|
920 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
|
921 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
922 def move_to_next_dataset(self): |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
923 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
|
924 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
|
925 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
|
926 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
|
927 else: |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
928 self.next_dataset_index +=1 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
929 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
|
930 self.next_dataset_index = 0 |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
931 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
|
932 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
|
933 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
934 def __iter__(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
935 return self |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
936 |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
937 def next(self): |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
938 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
|
939 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
|
940 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
|
941 extra_mb = [] |
|
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
942 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
|
943 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
|
944 extra_mb.append(self.next_iterator.next()) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
945 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
|
946 [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
|
947 [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
|
948 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
|
949 |
|
37
73c4212ba5b3
Factored the minibatch-writing code into an iterator class inside DataSet
bengioy@esprit.iro.umontreal.ca
parents:
36
diff
changeset
|
950 self.next_row+=minibatch_size |
|
38
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
951 self.next_dataset_row+=minibatch_size |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
952 if self.next_row+minibatch_size>len(dataset): |
|
d637ad8f7352
Finished first untested version of VStackedDataset
bengioy@esprit.iro.umontreal.ca
parents:
37
diff
changeset
|
953 self.move_to_next_dataset() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
954 return examples |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
955 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
|
956 |
| 41 | 957 class ArrayFieldsDataSet(DataSet): |
| 958 """ | |
| 959 Virtual super-class of datasets whose field values are numpy array, | |
| 960 thus defining valuesHStack and valuesVStack for sub-classes. | |
| 961 """ | |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
962 def __init__(self,description=None,field_types=None): |
|
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
963 DataSet.__init__(self,description,field_types) |
|
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
964 def valuesHStack(self,fieldnames,fieldvalues): |
| 41 | 965 """Concatenate field values horizontally, e.g. two vectors |
| 966 become a longer vector, two matrices become a wider matrix, etc.""" | |
| 967 return numpy.hstack(fieldvalues) | |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
968 def valuesVStack(self,fieldname,values): |
| 41 | 969 """Concatenate field values vertically, e.g. two vectors |
| 970 become a two-row matrix, two matrices become a longer matrix, etc.""" | |
| 971 return numpy.vstack(values) | |
| 972 | |
| 973 class ArrayDataSet(ArrayFieldsDataSet): | |
| 974 """ | |
| 975 An ArrayDataSet stores the fields as groups of columns in a numpy tensor, | |
| 976 whose first axis iterates over examples, second axis determines fields. | |
| 977 If the underlying array is N-dimensional (has N axes), then the field | |
| 978 values are (N-2)-dimensional objects (i.e. ordinary numbers if N=2). | |
| 979 """ | |
| 980 | |
|
188
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
981 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
|
982 """ |
|
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
983 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
|
984 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
|
985 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
|
986 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
|
987 """ |
|
188
f01ac276c6fb
added __contains__ to Dataset, added parent constructor call to ArrayDataSet
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
171
diff
changeset
|
988 ArrayFieldsDataSet.__init__(self, **kwargs) |
| 41 | 989 self.data=data_array |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
990 self.fields_columns=fields_columns |
| 41 | 991 |
| 992 # check consistency and complete slices definitions | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
993 for fieldname, fieldcolumns in self.fields_columns.items(): |
| 41 | 994 if type(fieldcolumns) is int: |
| 995 assert fieldcolumns>=0 and fieldcolumns<data_array.shape[1] | |
|
268
3f1cd8897fda
reverting dataset
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
266
diff
changeset
|
996 if 1: |
|
227
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
997 #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
|
998 # 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
|
999 # 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
|
1000 # - James 22/05/2008 |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1001 self.fields_columns[fieldname]=[fieldcolumns] |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1002 else: |
|
17c5d080964b
reinstating changeset 216 clobbered accidentally by 218
James Bergstra <bergstrj@iro.umontreal.ca>
parents:
223
diff
changeset
|
1003 self.fields_columns[fieldname]=fieldcolumns |
| 41 | 1004 elif type(fieldcolumns) is slice: |
| 1005 start,step=None,None | |
| 1006 if not fieldcolumns.start: | |
| 1007 start=0 | |
| 1008 if not fieldcolumns.step: | |
| 1009 step=1 | |
| 1010 if start or step: | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1011 self.fields_columns[fieldname]=slice(start,fieldcolumns.stop,step) |
| 41 | 1012 elif hasattr(fieldcolumns,"__iter__"): # something like a list |
| 1013 for i in fieldcolumns: | |
| 1014 assert i>=0 and i<data_array.shape[1] | |
| 1015 | |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1016 def fieldNames(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1017 return self.fields_columns.keys() |
| 41 | 1018 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1019 def __len__(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1020 return len(self.data) |
| 41 | 1021 |
| 80 | 1022 def __getitem__(self,key): |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1023 """More efficient implementation than the default __getitem__""" |
|
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1024 fieldnames=self.fields_columns.keys() |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1025 values=self.fields_columns.values() |
| 80 | 1026 if type(key) is int: |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1027 return Example(fieldnames, |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1028 [self.data[key,col] for col in values]) |
| 80 | 1029 if type(key) is slice: |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1030 return MinibatchDataSet(Example(fieldnames, |
|
243
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1031 [self.data[key,col] for col in values])) |
| 80 | 1032 if type(key) is list: |
| 1033 for i in range(len(key)): | |
| 1034 if self.hasFields(key[i]): | |
| 1035 key[i]=self.fields_columns[key[i]] | |
| 1036 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
|
1037 #we must separate differently for list as numpy |
| 128 | 1038 # 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
|
1039 # 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
|
1040 [self.data[key,:][:,col] |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1041 if isinstance(col,list) else |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1042 self.data[key,col] for col in values]), |
|
c8f19a9eb10f
Optimisation in ArrayDataSet::__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
242
diff
changeset
|
1043 |
| 128 | 1044 |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1045 self.valuesVStack,self.valuesHStack) |
| 80 | 1046 |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1047 # else check for a fieldname |
| 80 | 1048 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
|
1049 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
|
1050 # else we are trying to access a property of the dataset |
| 80 | 1051 assert key in self.__dict__ # else it means we are trying to access a non-existing property |
| 1052 return self.__dict__[key] | |
|
55
66619ce44497
Efficient implementation of getitem for ArrayDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
48
diff
changeset
|
1053 |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1054 def __iter__(self): |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1055 class ArrayDataSetIteratorIter(object): |
|
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1056 def __init__(self,dataset,fieldnames): |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1057 if fieldnames is None: fieldnames = dataset.fieldNames() |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1058 # store the resulting minibatch in a lookup-list of values |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1059 self.minibatch = LookupList(fieldnames,[0]*len(fieldnames)) |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1060 self.dataset=dataset |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1061 self.current=0 |
|
238
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1062 self.columns = [self.dataset.fields_columns[f] |
|
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1063 for f in self.minibatch._names] |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1064 self.l = self.dataset.data.shape[0] |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1065 def __iter__(self): |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1066 return self |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1067 def next(self): |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1068 #@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
|
1069 # Otherwise, MinibatchWrapAroundIterator do it. |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1070 if self.current>=self.l: |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1071 raise StopIteration |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1072 sub_data = self.dataset.data[self.current] |
|
238
ae1d85aca858
optimization in ArrayDataSet::__iter__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
235
diff
changeset
|
1073 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
|
1074 |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1075 self.current+=1 |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1076 return self.minibatch |
|
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1077 |
|
270
1cafd495098c
code cleanup and small optimisation
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
254
diff
changeset
|
1078 return ArrayDataSetIteratorIter(self,self.fieldNames()) |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1079 |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1080 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
|
1081 class ArrayDataSetIterator(object): |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1082 def __init__(self,dataset,fieldnames,minibatch_size,n_batches,offset): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1083 if fieldnames is None: fieldnames = dataset.fieldNames() |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
1084 # store the resulting minibatch in a lookup-list of values |
|
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
1085 self.minibatch = LookupList(fieldnames,[0]*len(fieldnames)) |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1086 self.dataset=dataset |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1087 self.minibatch_size=minibatch_size |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1088 assert offset>=0 and offset<len(dataset.data) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1089 assert offset+minibatch_size<=len(dataset.data) |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1090 self.current=offset |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1091 def __iter__(self): |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1092 return self |
|
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1093 def next(self): |
|
228
6f55e301c687
optimisation of ArrayDataSet
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
203
diff
changeset
|
1094 #@todo: we suppose that MinibatchWrapAroundIterator stop the iterator |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1095 sub_data = self.dataset.data[self.current:self.current+self.minibatch_size] |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
1096 self.minibatch._values = [sub_data[:,self.dataset.fields_columns[f]] for f in self.minibatch._names] |
|
43
e92244f30116
Corrected iterator logic errors
bengioy@grenat.iro.umontreal.ca
parents:
42
diff
changeset
|
1097 self.current+=self.minibatch_size |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
1098 return self.minibatch |
|
42
9b68774fcc6b
Testing basic functionality and removing obvious bugs
bengioy@grenat.iro.umontreal.ca
parents:
41
diff
changeset
|
1099 |
|
44
5a85fda9b19b
Fixed some more iterator bugs
bengioy@grenat.iro.umontreal.ca
parents:
43
diff
changeset
|
1100 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
|
1101 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1102 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1103 class CachedDataSet(DataSet): |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1104 """ |
| 167 | 1105 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
|
1106 (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
|
1107 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
|
1108 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
|
1109 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1110 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
|
1111 (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
|
1112 first access. |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1113 |
| 167 | 1114 @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
|
1115 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
|
1116 |
| 167 | 1117 @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
|
1118 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
|
1119 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
|
1120 """ |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1121 def __init__(self,source_dataset,cache_all_upon_construction=False): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1122 self.source_dataset=source_dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1123 self.cache_all_upon_construction=cache_all_upon_construction |
|
152
3f627e844cba
Fixes in CacheDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
151
diff
changeset
|
1124 self.cached_examples = [] |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1125 if cache_all_upon_construction: |
|
77
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1126 # this potentially brings all the source examples |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1127 # into memory at once, which may be too much |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1128 # the work could possibly be done by minibatches |
|
1e2bb5bad636
toying with different ways to implement learners
bengioy@bengiomac.local
parents:
74
diff
changeset
|
1129 # 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
|
1130 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
|
1131 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
|
1132 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
|
1133 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
|
1134 |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1135 self.fieldNames = source_dataset.fieldNames |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1136 self.hasFields = source_dataset.hasFields |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1137 self.valuesHStack = source_dataset.valuesHStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1138 self.valuesVStack = source_dataset.valuesVStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1139 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1140 def __len__(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1141 return len(self.source_dataset) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1142 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1143 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1144 class CacheIterator(object): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1145 def __init__(self,dataset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1146 self.dataset=dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1147 self.current=offset |
|
254
8ec867d12428
optimication in CachedDataSet.minibatches_nowrap
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
253
diff
changeset
|
1148 self.all_fields = self.dataset.fieldNames()==fieldnames |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1149 def __iter__(self): return self |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1150 def next(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1151 upper = self.current+minibatch_size |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1152 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
|
1153 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
|
1154 # cache everything from current length to upper |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1155 for example in self.dataset.source_dataset[cache_len:upper]: |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1156 self.dataset.cached_examples.append(example) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1157 all_fields_minibatch = Example(self.dataset.fieldNames(), |
|
152
3f627e844cba
Fixes in CacheDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
151
diff
changeset
|
1158 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
|
1159 self.current+=minibatch_size |
|
254
8ec867d12428
optimication in CachedDataSet.minibatches_nowrap
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
253
diff
changeset
|
1160 if self.all_fields: |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1161 return all_fields_minibatch |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1162 return Example(fieldnames,[all_fields_minibatch[name] for name in fieldnames]) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1163 return CacheIterator(self) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1164 |
| 153 | 1165 def __getitem__(self,i): |
| 1166 if type(i)==int and len(self.cached_examples)>i: | |
| 1167 return self.cached_examples[i] | |
| 1168 else: | |
|
251
7e6edee187e3
optimization of CachedDataSet__getitem__
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
243
diff
changeset
|
1169 return self.source_dataset[i] |
|
252
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1170 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1171 def __iter__(self): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1172 class CacheIteratorIter(object): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1173 def __init__(self,dataset): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1174 self.dataset=dataset |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1175 self.l = len(dataset) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1176 self.current = 0 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1177 self.fieldnames = self.dataset.fieldNames() |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1178 self.example = LookupList(self.fieldnames,[0]*len(self.fieldnames)) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1179 def __iter__(self): return self |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1180 def next(self): |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1181 if self.current>=self.l: |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1182 raise StopIteration |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1183 cache_len = len(self.dataset.cached_examples) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1184 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
|
1185 # cache everything from current length to upper |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1186 self.dataset.cached_examples.append( |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1187 self.dataset.source_dataset[self.current]) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1188 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
|
1189 self.current+=1 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1190 return self.example |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1191 |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1192 return CacheIteratorIter(self) |
|
856d14dc4468
implemented CachedDataSet.__iter__ as an optimization
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
251
diff
changeset
|
1193 |
|
57
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1194 class ApplyFunctionDataSet(DataSet): |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1195 """ |
| 167 | 1196 A L{DataSet} that contains as fields the results of applying a |
| 1197 given function example-wise or minibatch-wise to all the fields of | |
| 1198 an input dataset. The output of the function should be an iterable | |
| 1199 (e.g. a list or a LookupList) over the resulting values. | |
|
197
96d891448107
more explicit comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
196
diff
changeset
|
1200 |
|
96d891448107
more explicit comment
Frederic Bastien <bastienf@iro.umontreal.ca>
parents:
196
diff
changeset
|
1201 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
|
1202 |
| 167 | 1203 In minibatch mode, the function is expected to work on minibatches |
| 1204 (takes a minibatch in input and returns a minibatch in output). More | |
| 1205 precisely, it means that each element of the input or output list | |
| 1206 should be iterable and indexable over the individual example values | |
| 1207 (typically these elements will be numpy arrays). All of the elements | |
| 1208 in the input and output lists should have the same length, which is | |
| 1209 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
|
1210 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1211 The function is applied each time an example or a minibatch is accessed. |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1212 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
|
1213 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1214 If the values_{h,v}stack functions are not provided, then |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1215 the input_dataset.values{H,V}Stack functions are used by default. |
|
57
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1216 """ |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1217 def __init__(self,input_dataset,function,output_names,minibatch_mode=True, |
|
74
b4159cbdc06b
Fixed errors raised by test_dataset
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
73
diff
changeset
|
1218 values_hstack=None,values_vstack=None, |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1219 description=None,fieldtypes=None): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1220 """ |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1221 Constructor takes an input dataset that has as many fields as the function |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1222 expects as inputs. The resulting dataset has as many fields as the function |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1223 produces as outputs, and that should correspond to the number of output names |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1224 (provided in a list). |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1225 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1226 Note that the expected semantics of the function differs in minibatch mode |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1227 (it takes minibatches of inputs and produces minibatches of outputs, as |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1228 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
|
1229 |
|
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
|
1230 TBM: are filedtypes the old field types (from input_dataset) or the new ones |
|
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
|
1231 (for the new dataset created)? |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1232 """ |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1233 self.input_dataset=input_dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1234 self.function=function |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1235 self.output_names=output_names |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1236 self.minibatch_mode=minibatch_mode |
|
134
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1237 DataSet.__init__(self,description,fieldtypes) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1238 self.valuesHStack = values_hstack if values_hstack else input_dataset.valuesHStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1239 self.valuesVStack = values_vstack if values_vstack else input_dataset.valuesVStack |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1240 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1241 def __len__(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1242 return len(self.input_dataset) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1243 |
|
134
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1244 def fieldNames(self): |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1245 return self.output_names |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1246 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1247 def minibatches_nowrap(self,fieldnames,minibatch_size,n_batches,offset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1248 class ApplyFunctionIterator(object): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1249 def __init__(self,output_dataset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1250 self.input_dataset=output_dataset.input_dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1251 self.output_dataset=output_dataset |
|
134
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1252 self.input_iterator=self.input_dataset.minibatches(minibatch_size=minibatch_size, |
|
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1253 n_batches=n_batches,offset=offset).__iter__() |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1254 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1255 def __iter__(self): return self |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1256 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1257 def next(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1258 function_inputs = self.input_iterator.next() |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1259 all_output_names = self.output_dataset.output_names |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1260 if self.output_dataset.minibatch_mode: |
|
134
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1261 function_outputs = self.output_dataset.function(*function_inputs) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1262 else: |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1263 input_examples = zip(*function_inputs) |
| 201 | 1264 output_examples = [self.output_dataset.function(*input_example) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1265 for input_example in input_examples] |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1266 function_outputs = [self.output_dataset.valuesVStack(name,values) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1267 for name,values in zip(all_output_names, |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1268 zip(*output_examples))] |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1269 all_outputs = Example(all_output_names,function_outputs) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1270 if fieldnames==all_output_names: |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1271 return all_outputs |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1272 return Example(fieldnames,[all_outputs[name] for name in fieldnames]) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1273 |
|
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
|
1274 |
|
134
3f4e5c9bdc5e
Fixes to ApplyFunctionDataSet and other things to make learner and mlp work
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
132
diff
changeset
|
1275 return ApplyFunctionIterator(self) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1276 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1277 def __iter__(self): # only implemented for increased efficiency |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1278 class ApplyFunctionSingleExampleIterator(object): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1279 def __init__(self,output_dataset): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1280 self.current=0 |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1281 self.output_dataset=output_dataset |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1282 self.input_iterator=output_dataset.input_dataset.__iter__() |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1283 def __iter__(self): return self |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1284 def next(self): |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1285 if self.output_dataset.minibatch_mode: |
| 200 | 1286 function_inputs = [[input] for input in self.input_iterator.next()] |
| 199 | 1287 outputs = self.output_dataset.function(*function_inputs) |
| 1288 assert all([hasattr(output,'__iter__') for output in outputs]) | |
| 1289 function_outputs = [output[0] for output in outputs] | |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1290 else: |
| 200 | 1291 function_inputs = self.input_iterator.next() |
| 199 | 1292 function_outputs = self.output_dataset.function(*function_inputs) |
|
73
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1293 return Example(self.output_dataset.output_names,function_outputs) |
|
69f97aad3faf
Coded untested ApplyFunctionDataSet and CacheDataSet
bengioy@bengiomac.local
parents:
72
diff
changeset
|
1294 return ApplyFunctionSingleExampleIterator(self) |
|
57
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1295 |
|
1aabd2e2bb5f
Added empty classes with doc: CachedDataSet and ApplyFunctionDataSet
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
56
diff
changeset
|
1296 |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1297 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
|
1298 """ |
| 167 | 1299 Wraps an arbitrary L{DataSet} into one for supervised learning tasks |
| 1300 by forcing the user to define a set of fields as the 'input' field | |
| 1301 and a set of fields as the 'target' field. Optionally, a single | |
| 1302 weight_field can also be defined. | |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1303 """ |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1304 args = ((input_fields,'input'),(output_fields,'target')) |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1305 if weight_field: args+=(([weight_field],'weight')) |
|
36
438440ba0627
Rewriting dataset.py completely
bengioy@zircon.iro.umontreal.ca
parents:
29
diff
changeset
|
1306 return src_dataset.merge_fields(*args) |
|
23
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1307 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1308 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1309 |
|
526e192b0699
Working on ApplyFunctionDataSet, added constraint that
bengioy@esprit.iro.umontreal.ca
parents:
22
diff
changeset
|
1310 |