Mercurial > pylearn

annotate denoising_aa.py @ 223:517364d48ae0

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
should have solved the problem with minibatches not handling subsets of fieldnames, although maybe not super efficient
author Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
date Fri, 23 May 2008 16:01:01 -0400
parents df3fae88ab46
children 9e96fe8b955c
rev   line source
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
1 """
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
2 A denoising auto-encoder
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
3 """
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
4
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
5 import theano
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
6 from theano.formula import *
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
7 from learner import *
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
8 from theano import tensor as t
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
9 from nnet_ops import *
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
10 import math
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
11 from misc import *
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
12 from theano.tensor_random import binomial
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
13
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
14 def hiding_corruption_formula(seed,average_fraction_hidden):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
15 """
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
16 Return a formula for the corruption process, in which a random
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
17 subset of the input numbers are hidden (mapped to 0).
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
18
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
19 @param seed: seed of the random generator
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
20 @type seed: anything that numpy.random.RandomState accepts
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
21
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
22 @param average_fraction_hidden: the probability with which each
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
23 input number is hidden (set to 0).
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
24 @type average_fraction_hidden: 0 <= real number <= 1
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
25 """
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
26 class HidingCorruptionFormula(Formulas):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
27 x = t.matrix()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
28 corrupted_x = x * binomial(seed,x,1,fraction_sampled)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
29
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
30 return HidingCorruptionFormula()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
31
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
32 def squash_affine_formula(squash_function=sigmoid):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
33 """
218
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
34 Simply does: squash_function(b + xW)
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
35 By convention prefix the parameters by _
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
36 """
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
37 class SquashAffineFormula(Formulas):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
38 x = t.matrix() # of dimensions minibatch_size x n_inputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
39 _b = t.row() # of dimensions 1 x n_outputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
40 _W = t.matrix() # of dimensions n_inputs x n_outputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
41 a = _b + t.dot(x,_W) # of dimensions minibatch_size x n_outputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
42 y = squash_function(a)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
43 return SquashAffineFormula()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
44
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
45 def gradient_descent_update_formula():
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
46 class GradientDescentUpdateFormula(Formula):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
47 param = t.matrix()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
48 learning_rate = t.scalar()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
49 cost = t.column() # cost of each example in a minibatch
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
50 param_update = t.add_inplace(param, -learning_rate*t.sgrad(cost))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
51 return gradient_descent_update_formula()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
52
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
53 def probabilistic_classifier_loss_formula():
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
54 class ProbabilisticClassifierLossFormula(Formulas):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
55 a = t.matrix() # of dimensions minibatch_size x n_classes, pre-softmax output
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
56 target_class = t.ivector() # dimension (minibatch_size)
218
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
57 nll, probability_predictions = crossentropy_softmax_1hot(a, target_class) # defined in nnet_ops.py
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
58 return ProbabilisticClassifierLossFormula()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
59
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
60 def binomial_cross_entropy_formula():
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
61 class BinomialCrossEntropyFormula(Formulas):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
62 a = t.matrix() # pre-sigmoid activations, minibatch_size x dim
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
63 p = sigmoid(a) # model prediction
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
64 q = t.matrix() # target binomial probabilities, minibatch_size x dim
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
65 # using the identity softplus(a) - softplus(-a) = a,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
66 # we obtain that q log(p) + (1-q) log(1-p) = q a - softplus(a)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
67 nll = -t.sum(q*a - softplus(-a))
218
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
68 # next line was missing... hope it's all correct above
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
69 return BinomialCrossEntropyFormula()
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
70
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
71 def squash_affine_autoencoder_formula(hidden_squash=t.tanh,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
72 reconstruction_squash=sigmoid,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
73 share_weights=True,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
74 reconstruction_nll_formula=binomial_cross_entropy_formula(),
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
75 update_formula=gradient_descent_update_formula):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
76 if share_weights:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
77 autoencoder = squash_affine_formula(hidden_squash).rename(a='code_a') + \
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
78 squash_affine_formula(reconstruction_squash).rename(x='hidden',y='reconstruction',_b='_c') + \
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
79 reconstruction_nll_formula
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
80 else:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
81 autoencoder = squash_affine_formula(hidden_squash).rename(a='code_a',_W='_W1') + \
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
82 squash_affine_formula(reconstruction_squash).rename(x='hidden',y='reconstruction',_b='_c',_W='_W2') + \
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
83 reconstruction_nll_formula
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
84 autoencoder = autoencoder + [update_formula().rename(cost = 'nll',
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
85 param = p)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
86 for p in autoencoder.get_all('_.*')]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
87 return autoencoder
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
88
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
89
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
90 # @todo: try other corruption formulae. The above is the default one.
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
91 # not quite used in the ICML paper... (had a fixed number of 0s).
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
92
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
93 class DenoisingAutoEncoder(LearningAlgorithm):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
94
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
95 def __init__(self,n_inputs,n_hidden_per_layer,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
96 learning_rate=0.1,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
97 max_n_epochs=100,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
98 L1_regularizer=0,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
99 init_range=1.,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
100 corruption_formula = hiding_corruption_formula(),
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
101 autoencoder = squash_affine_autoencoder_formula(),
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
102 minibatch_size=None,linker = "c|py"):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
103 for name,val in locals().items():
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
104 if val is not self: self.__setattribute__(name,val)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
105 self.denoising_autoencoder_formula = corruption_formula + autoencoder.rename(x='corrupted_x')
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
106
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
107 def __call__(self, training_set=None):
218
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
108 """ Allocate and optionnaly train a model"""
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
109 model = DenoisingAutoEncoderModel(self)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
110 if training_set:
218
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
111 print 'DenoisingAutoEncoder(): what do I do if training_set????'
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
112 # copied from mlp_factory_approach:
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
113 if len(trainset) == sys.maxint:
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
114 raise NotImplementedError('Learning from infinite streams is not supported')
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
115 nval = int(self.validation_portion * len(trainset))
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
116 nmin = len(trainset) - nval
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
117 assert nmin >= 0
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
118 minset = trainset[:nmin] #real training set for minimizing loss
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
119 valset = trainset[nmin:] #validation set for early stopping
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
120 best = model
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
121 for stp in self.early_stopper():
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
122 model.update(
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
123 minset.minibatches([input, target], minibatch_size=min(32,
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
124 len(trainset))))
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
125 #print 'mlp.__call__(), we did an update'
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
126 if stp.set_score:
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
127 stp.score = model(valset, ['loss_01'])
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
128 if (stp.score < stp.best_score):
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
129 best = copy.copy(model)
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
130 model = best
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
131 # end of the copy from mlp_factory_approach
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
132
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
133 return model
df3fae88ab46 small debugging
Thierry Bertin-Mahieux <bertinmt@iro.umontreal.ca>
parents: 211
diff changeset
134
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
135
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
136 def compile(self, inputs, outputs):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
137 return theano.function(inputs,outputs,unpack_single=False,linker=self.linker)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
138
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
139 class DenoisingAutoEncoderModel(LearnerModel):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
140 def __init__(self,learning_algorithm,params):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
141 self.learning_algorithm=learning_algorithm
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
142 self.params=params
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
143 v = learning_algorithm.v
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
144 self.update_fn = learning_algorithm.compile(learning_algorithm.denoising_autoencoder_formula.inputs,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
145 learning_algorithm.denoising_autoencoder_formula.outputs)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
146
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
147 def update(self, training_set, train_stats_collector=None):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
148
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: 210
diff changeset
149 print 'dont update you crazy frog!'
210
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
150
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
151 # old stuff
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
152
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
153 # self._learning_rate = t.scalar('learning_rate') # this is the symbol
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
154 # self.L1_regularizer = L1_regularizer
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
155 # self._L1_regularizer = t.scalar('L1_regularizer')
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
156 # self._input = t.matrix('input') # n_examples x n_inputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
157 # self._W = t.matrix('W')
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
158 # self._b = t.row('b')
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
159 # self._c = t.row('b')
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
160 # self._regularization_term = self._L1_regularizer * t.sum(t.abs(self._W))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
161 # self._corrupted_input = corruption_process(self._input)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
162 # self._hidden = t.tanh(self._b + t.dot(self._input, self._W.T))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
163 # self._reconstruction_activations =self._c+t.dot(self._hidden,self._W)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
164 # self._nll,self._output = crossentropy_softmax_1hot(Print("output_activations")(self._output_activations),self._target_vector)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
165 # self._output_class = t.argmax(self._output,1)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
166 # self._class_error = t.neq(self._output_class,self._target_vector)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
167 # self._minibatch_criterion = self._nll + self._regularization_term / t.shape(self._input)[0]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
168 # OnlineGradientTLearner.__init__(self)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
169
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
170 # def attributeNames(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
171 # return ["parameters","b1","W2","b2","W2", "L2_regularizer","regularization_term"]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
172
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
173 # def parameterAttributes(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
174 # return ["b1","W1", "b2", "W2"]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
175
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
176 # def updateMinibatchInputFields(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
177 # return ["input","target"]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
178
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
179 # def updateEndOutputAttributes(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
180 # return ["regularization_term"]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
181
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
182 # def lossAttribute(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
183 # return "minibatch_criterion"
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
184
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
185 # def defaultOutputFields(self, input_fields):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
186 # output_fields = ["output", "output_class",]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
187 # if "target" in input_fields:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
188 # output_fields += ["class_error", "nll"]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
189 # return output_fields
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
190
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
191 # def allocate(self,minibatch):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
192 # minibatch_n_inputs = minibatch["input"].shape[1]
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
193 # if not self._n_inputs:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
194 # self._n_inputs = minibatch_n_inputs
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
195 # self.b1 = numpy.zeros((1,self._n_hidden))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
196 # self.b2 = numpy.zeros((1,self._n_outputs))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
197 # self.forget()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
198 # elif self._n_inputs!=minibatch_n_inputs:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
199 # # if the input changes dimension on the fly, we resize and forget everything
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
200 # self.forget()
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
201
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
202 # def forget(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
203 # if self._n_inputs:
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
204 # r = self._init_range/math.sqrt(self._n_inputs)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
205 # self.W1 = numpy.random.uniform(low=-r,high=r,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
206 # size=(self._n_hidden,self._n_inputs))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
207 # r = self._init_range/math.sqrt(self._n_hidden)
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
208 # self.W2 = numpy.random.uniform(low=-r,high=r,
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
209 # size=(self._n_outputs,self._n_hidden))
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
210 # self.b1[:]=0
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
211 # self.b2[:]=0
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
212 # self._n_epochs=0
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
213
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
214 # def isLastEpoch(self):
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
215 # self._n_epochs +=1
ffd50efefb70 work in progress denoising auto-encoder
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
216 # return self._n_epochs>=self._max_n_epochs