Mercurial > ift6266

annotate baseline/mlp/mlp_nist.py @ 368:d391ad815d89

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Correction d'un bug avec la fonction de log-likelihood pour utilisation de non-linearite de sortie sigmoides
author SylvainPL <sylvain.pannetier.lebeuf@umontreal.ca>
date Fri, 23 Apr 2010 12:12:03 -0400
parents 76b7182dd32e
children 60a4432b8071
rev   line source
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
1 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
2 This tutorial introduces the multilayer perceptron using Theano.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
3
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
4 A multilayer perceptron is a logistic regressor where
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
5 instead of feeding the input to the logistic regression you insert a
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
6 intermidiate layer, called the hidden layer, that has a nonlinear
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
7 activation function (usually tanh or sigmoid) . One can use many such
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
8 hidden layers making the architecture deep. The tutorial will also tackle
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
9 the problem of MNIST digit classification.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
10
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
11 .. math::
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
12
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
13 f(x) = G( b^{(2)} + W^{(2)}( s( b^{(1)} + W^{(1)} x))),
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
14
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
15 References:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
16
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
17 - textbooks: "Pattern Recognition and Machine Learning" -
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
18 Christopher M. Bishop, section 5
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
19
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
20 TODO: recommended preprocessing, lr ranges, regularization ranges (explain
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
21 to do lr first, then add regularization)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
22
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
23 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
24 __docformat__ = 'restructedtext en'
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
25
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
26 import sys
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
27 import pdb
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
28 import numpy
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
29 import pylab
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
30 import theano
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
31 import theano.tensor as T
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
32 import time
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
33 import theano.tensor.nnet
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
34 import pylearn
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
35 import theano,pylearn.version,ift6266
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
36 from pylearn.io import filetensor as ft
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
37 from ift6266 import datasets
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
38
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
39 data_path = '/data/lisa/data/nist/by_class/'
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
40
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
41 class MLP(object):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
42 """Multi-Layer Perceptron Class
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
43
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
44 A multilayer perceptron is a feedforward artificial neural network model
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
45 that has one layer or more of hidden units and nonlinear activations.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
46 Intermidiate layers usually have as activation function thanh or the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
47 sigmoid function while the top layer is a softamx layer.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
48 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
49
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
50
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
51
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
52 def __init__(self, input, n_in, n_hidden, n_out,learning_rate):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
53 """Initialize the parameters for the multilayer perceptron
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
54
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
55 :param input: symbolic variable that describes the input of the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
56 architecture (one minibatch)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
57
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
58 :param n_in: number of input units, the dimension of the space in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
59 which the datapoints lie
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
60
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
61 :param n_hidden: number of hidden units
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
62
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
63 :param n_out: number of output units, the dimension of the space in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
64 which the labels lie
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
65
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
66 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
67
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
68 # initialize the parameters theta = (W1,b1,W2,b2) ; note that this
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
69 # example contains only one hidden layer, but one can have as many
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
70 # layers as he/she wishes, making the network deeper. The only
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
71 # problem making the network deep this way is during learning,
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
72 # backpropagation being unable to move the network from the starting
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
73 # point towards; this is where pre-training helps, giving a good
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
74 # starting point for backpropagation, but more about this in the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
75 # other tutorials
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
76
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
77 # `W1` is initialized with `W1_values` which is uniformely sampled
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
78 # from -6./sqrt(n_in+n_hidden) and 6./sqrt(n_in+n_hidden)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
79 # the output of uniform if converted using asarray to dtype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
80 # theano.config.floatX so that the code is runable on GPU
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
81 W1_values = numpy.asarray( numpy.random.uniform( \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
82 low = -numpy.sqrt(6./(n_in+n_hidden)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
83 high = numpy.sqrt(6./(n_in+n_hidden)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
84 size = (n_in, n_hidden)), dtype = theano.config.floatX)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
85 # `W2` is initialized with `W2_values` which is uniformely sampled
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
86 # from -6./sqrt(n_hidden+n_out) and 6./sqrt(n_hidden+n_out)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
87 # the output of uniform if converted using asarray to dtype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
88 # theano.config.floatX so that the code is runable on GPU
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
89 W2_values = numpy.asarray( numpy.random.uniform(
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
90 low = -numpy.sqrt(6./(n_hidden+n_out)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
91 high= numpy.sqrt(6./(n_hidden+n_out)),\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
92 size= (n_hidden, n_out)), dtype = theano.config.floatX)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
93
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
94 self.W1 = theano.shared( value = W1_values )
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
95 self.b1 = theano.shared( value = numpy.zeros((n_hidden,),
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
96 dtype= theano.config.floatX))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
97 self.W2 = theano.shared( value = W2_values )
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
98 self.b2 = theano.shared( value = numpy.zeros((n_out,),
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
99 dtype= theano.config.floatX))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
100
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
101 #include the learning rate in the classifer so
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
102 #we can modify it on the fly when we want
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
103 lr_value=learning_rate
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
104 self.lr=theano.shared(value=lr_value)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
105 # symbolic expression computing the values of the hidden layer
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
106 self.hidden = T.tanh(T.dot(input, self.W1)+ self.b1)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
107
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
108
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
109
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
110 # symbolic expression computing the values of the top layer
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
111 self.p_y_given_x= T.nnet.softmax(T.dot(self.hidden, self.W2)+self.b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
112
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
113 # compute prediction as class whose probability is maximal in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
114 # symbolic form
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
115 self.y_pred = T.argmax( self.p_y_given_x, axis =1)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
116 self.y_pred_num = T.argmax( self.p_y_given_x[0:9], axis =1)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
117
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
118
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
119
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
120
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
121 # L1 norm ; one regularization option is to enforce L1 norm to
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
122 # be small
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
123 self.L1 = abs(self.W1).sum() + abs(self.W2).sum()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
124
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
125 # square of L2 norm ; one regularization option is to enforce
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
126 # square of L2 norm to be small
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
127 self.L2_sqr = (self.W1**2).sum() + (self.W2**2).sum()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
128
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
129
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
130
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
131 def negative_log_likelihood(self, y):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
132 """Return the mean of the negative log-likelihood of the prediction
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
133 of this model under a given target distribution.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
134
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
135 .. math::
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
136
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
137 \frac{1}{|\mathcal{D}|}\mathcal{L} (\theta=\{W,b\}, \mathcal{D}) =
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
138 \frac{1}{|\mathcal{D}|}\sum_{i=0}^{|\mathcal{D}|} \log(P(Y=y^{(i)}|x^{(i)}, W,b)) \\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
139 \ell (\theta=\{W,b\}, \mathcal{D})
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
140
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
141
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
142 :param y: corresponds to a vector that gives for each example the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
143 :correct label
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
144 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
145 return -T.mean(T.log(self.p_y_given_x)[T.arange(y.shape[0]),y])
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
146
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
147
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
148
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
149
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
150 def errors(self, y):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
151 """Return a float representing the number of errors in the minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
152 over the total number of examples of the minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
153 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
154
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
155 # check if y has same dimension of y_pred
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
156 if y.ndim != self.y_pred.ndim:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
157 raise TypeError('y should have the same shape as self.y_pred',
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
158 ('y', target.type, 'y_pred', self.y_pred.type))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
159 # check if y is of the correct datatype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
160 if y.dtype.startswith('int'):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
161 # the T.neq operator returns a vector of 0s and 1s, where 1
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
162 # represents a mistake in prediction
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
163 return T.mean(T.neq(self.y_pred, y))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
164 else:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
165 raise NotImplementedError()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
166
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
167 def mlp_get_nist_error(model_name='/u/mullerx/ift6266h10_sandbox_db/xvm_final_lr1_p073/8/best_model.npy.npz',
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
168 data_set=0):
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
169
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
170
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
171
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
172 # allocate symbolic variables for the data
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
173 x = T.fmatrix() # the data is presented as rasterized images
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
174 y = T.lvector() # the labels are presented as 1D vector of
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
175 # [long int] labels
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
176
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
177 # load the data set and create an mlp based on the dimensions of the model
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
178 model=numpy.load(model_name)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
179 W1=model['W1']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
180 W2=model['W2']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
181 b1=model['b1']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
182 b2=model['b2']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
183 nb_hidden=b1.shape[0]
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
184 input_dim=W1.shape[0]
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
185 nb_targets=b2.shape[0]
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
186 learning_rate=0.1
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
187
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
188
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
189 if data_set==0:
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
190 dataset=datasets.nist_all()
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
191 elif data_set==1:
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
192 dataset=datasets.nist_P07()
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
193
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
194
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
195 classifier = MLP( input=x,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
196 n_in=input_dim,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
197 n_hidden=nb_hidden,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
198 n_out=nb_targets,
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
199 learning_rate=learning_rate)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
200
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
201
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
202 #overwrite weights with weigths from model
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
203 classifier.W1.value=W1
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
204 classifier.W2.value=W2
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
205 classifier.b1.value=b1
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
206 classifier.b2.value=b2
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
207
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
208
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
209 cost = classifier.negative_log_likelihood(y) \
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
210 + 0.0 * classifier.L1 \
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
211 + 0.0 * classifier.L2_sqr
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
212
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
213 # compiling a theano function that computes the mistakes that are made by
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
214 # the model on a minibatch
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
215 test_model = theano.function([x,y], classifier.errors(y))
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
216
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
217
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
218
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
219 #get the test error
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
220 #use a batch size of 1 so we can get the sub-class error
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
221 #without messing with matrices (will be upgraded later)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
222 test_score=0
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
223 temp=0
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
224 for xt,yt in dataset.test(20):
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
225 test_score += test_model(xt,yt)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
226 temp = temp+1
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
227 test_score /= temp
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
228
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
229
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
230 return test_score*100
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
231
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
232
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
233
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
234
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
235
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
236
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
237 def mlp_full_nist( verbose = 1,\
145
8ceaaf812891 changed adaptive lr flag from bool to int for jobman issues
XavierMuller
parents: 143
diff changeset
238 adaptive_lr = 0,\
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
239 data_set=0,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
240 learning_rate=0.01,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
241 L1_reg = 0.00,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
242 L2_reg = 0.0001,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
243 nb_max_exemples=1000000,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
244 batch_size=20,\
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
245 nb_hidden = 30,\
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
246 nb_targets = 62,
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
247 tau=1e6,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
248 lr_t2_factor=0.5,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
249 init_model=0,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
250 channel=0):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
251
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
252
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
253 if channel!=0:
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
254 channel.save()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
255 configuration = [learning_rate,nb_max_exemples,nb_hidden,adaptive_lr]
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
256
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
257 #save initial learning rate if classical adaptive lr is used
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
258 initial_lr=learning_rate
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
259 max_div_count=1000
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
260
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
261
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
262 total_validation_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
263 total_train_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
264 learning_rate_list=[]
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
265 best_training_error=float('inf');
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
266 divergence_flag_list=[]
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
267
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
268 if data_set==0:
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
269 dataset=datasets.nist_all()
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
270 elif data_set==1:
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
271 dataset=datasets.nist_P07()
349
22efb4968054 added pnist support, will check in code for data set iterator later
xaviermuller
parents: 338
diff changeset
272 elif data_set==2:
22efb4968054 added pnist support, will check in code for data set iterator later
xaviermuller
parents: 338
diff changeset
273 dataset=datasets.PNIST07()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
274
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
275
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
276
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
277
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
278 ishape = (32,32) # this is the size of NIST images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
279
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
280 # allocate symbolic variables for the data
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
281 x = T.fmatrix() # the data is presented as rasterized images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
282 y = T.lvector() # the labels are presented as 1D vector of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
283 # [long int] labels
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
284
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
285
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
286 # construct the logistic regression class
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
287 classifier = MLP( input=x,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
288 n_in=32*32,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
289 n_hidden=nb_hidden,\
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
290 n_out=nb_targets,
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
291 learning_rate=learning_rate)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
292
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
293
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
294 # check if we want to initialise the weights with a previously calculated model
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
295 # dimensions must be consistent between old model and current configuration!!!!!! (nb_hidden and nb_targets)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
296 if init_model!=0:
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
297 old_model=numpy.load(init_model)
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
298 classifier.W1.value=old_model['W1']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
299 classifier.W2.value=old_model['W2']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
300 classifier.b1.value=old_model['b1']
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
301 classifier.b2.value=old_model['b2']
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
302
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
303
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
304 # the cost we minimize during training is the negative log likelihood of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
305 # the model plus the regularization terms (L1 and L2); cost is expressed
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
306 # here symbolically
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
307 cost = classifier.negative_log_likelihood(y) \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
308 + L1_reg * classifier.L1 \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
309 + L2_reg * classifier.L2_sqr
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
310
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
311 # compiling a theano function that computes the mistakes that are made by
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
312 # the model on a minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
313 test_model = theano.function([x,y], classifier.errors(y))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
314
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
315 # compute the gradient of cost with respect to theta = (W1, b1, W2, b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
316 g_W1 = T.grad(cost, classifier.W1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
317 g_b1 = T.grad(cost, classifier.b1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
318 g_W2 = T.grad(cost, classifier.W2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
319 g_b2 = T.grad(cost, classifier.b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
320
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
321 # specify how to update the parameters of the model as a dictionary
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
322 updates = \
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
323 { classifier.W1: classifier.W1 - classifier.lr*g_W1 \
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
324 , classifier.b1: classifier.b1 - classifier.lr*g_b1 \
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
325 , classifier.W2: classifier.W2 - classifier.lr*g_W2 \
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
326 , classifier.b2: classifier.b2 - classifier.lr*g_b2 }
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
327
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
328 # compiling a theano function `train_model` that returns the cost, but in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
329 # the same time updates the parameter of the model based on the rules
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
330 # defined in `updates`
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
331 train_model = theano.function([x, y], cost, updates = updates )
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
332
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
333
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
334
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
335
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
336
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
337
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
338
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
339
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
340
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
341 #conditions for stopping the adaptation:
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
342 #1) we have reached nb_max_exemples (this is rounded up to be a multiple of the train size so we always do at least 1 epoch)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
343 #2) validation error is going up twice in a row(probable overfitting)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
344
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
345 # This means we no longer stop on slow convergence as low learning rates stopped
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
346 # too fast but instead we will wait for the valid error going up 3 times in a row
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
347 # We save the curb of the validation error so we can always go back to check on it
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
348 # and we save the absolute best model anyway, so we might as well explore
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
349 # a bit when diverging
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
350
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
351 #approximate number of samples in the nist training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
352 #this is just to have a validation frequency
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
353 #roughly proportionnal to the original nist training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
354 n_minibatches = 650000/batch_size
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
355
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
356
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
357 patience =2*nb_max_exemples/batch_size #in units of minibatch
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
358 validation_frequency = n_minibatches/4
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
359
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
360
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
361
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
362
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
363
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
364 best_validation_loss = float('inf')
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
365 best_iter = 0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
366 test_score = 0.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
367 start_time = time.clock()
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
368 time_n=0 #in unit of exemples
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
369 minibatch_index=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
370 epoch=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
371 temp=0
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
372 divergence_flag=0
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
373
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
374
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
375
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
376
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
377 print 'starting training'
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
378 sys.stdout.flush()
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
379 while(minibatch_index*batch_size<nb_max_exemples):
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
380
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
381 for x, y in dataset.train(batch_size):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
382
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
383 #if we are using the classic learning rate deacay, adjust it before training of current mini-batch
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
384 if adaptive_lr==2:
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
385 classifier.lr.value = tau*initial_lr/(tau+time_n)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
386
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
387
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
388 #train model
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
389 cost_ij = train_model(x,y)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
390
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
391 if (minibatch_index) % validation_frequency == 0:
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
392 #save the current learning rate
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
393 learning_rate_list.append(classifier.lr.value)
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
394 divergence_flag_list.append(divergence_flag)
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
395
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
396
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
397
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
398 # compute the validation error
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
399 this_validation_loss = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
400 temp=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
401 for xv,yv in dataset.valid(1):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
402 # sum up the errors for each minibatch
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
403 this_validation_loss += test_model(xv,yv)
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
404 temp=temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
405 # get the average by dividing with the number of minibatches
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
406 this_validation_loss /= temp
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
407 #save the validation loss
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
408 total_validation_error_list.append(this_validation_loss)
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
409
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
410 print(('epoch %i, minibatch %i, learning rate %f current validation error %f ') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
411 (epoch, minibatch_index+1,classifier.lr.value,
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
412 this_validation_loss*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
413 sys.stdout.flush()
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
414
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
415 #save temp results to check during training
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
416 numpy.savez('temp_results.npy',config=configuration,total_validation_error_list=total_validation_error_list,\
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
417 learning_rate_list=learning_rate_list, divergence_flag_list=divergence_flag_list)
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
418
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
419 # if we got the best validation score until now
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
420 if this_validation_loss < best_validation_loss:
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
421 # save best validation score and iteration number
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
422 best_validation_loss = this_validation_loss
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
423 best_iter = minibatch_index
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
424 #reset divergence flag
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
425 divergence_flag=0
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
426
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
427 #save the best model. Overwrite the current saved best model so
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
428 #we only keep the best
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
429 numpy.savez('best_model.npy', config=configuration, W1=classifier.W1.value, W2=classifier.W2.value, b1=classifier.b1.value,\
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
430 b2=classifier.b2.value, minibatch_index=minibatch_index)
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
431
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
432 # test it on the test set
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
433 test_score = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
434 temp =0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
435 for xt,yt in dataset.test(batch_size):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
436 test_score += test_model(xt,yt)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
437 temp = temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
438 test_score /= temp
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
439
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
440 print(('epoch %i, minibatch %i, test error of best '
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
441 'model %f %%') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
442 (epoch, minibatch_index+1,
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
443 test_score*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
444 sys.stdout.flush()
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
445
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
446 # if the validation error is going up, we are overfitting (or oscillating)
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
447 # check if we are allowed to continue and if we will adjust the learning rate
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
448 elif this_validation_loss >= best_validation_loss:
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
449
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
450
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
451 # In non-classic learning rate decay, we modify the weight only when
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
452 # validation error is going up
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
453 if adaptive_lr==1:
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
454 classifier.lr.value=classifier.lr.value*lr_t2_factor
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
455
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
456
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
457 #cap the patience so we are allowed to diverge max_div_count times
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
458 #if we are going up max_div_count in a row, we will stop immediatelty by modifying the patience
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
459 divergence_flag = divergence_flag +1
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
460
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
461
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
462 #calculate the test error at this point and exit
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
463 # test it on the test set
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
464 test_score = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
465 temp=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
466 for xt,yt in dataset.test(batch_size):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
467 test_score += test_model(xt,yt)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
468 temp=temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
469 test_score /= temp
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
470
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
471 print ' validation error is going up, possibly stopping soon'
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
472 print((' epoch %i, minibatch %i, test error of best '
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
473 'model %f %%') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
474 (epoch, minibatch_index+1,
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
475 test_score*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
476 sys.stdout.flush()
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
477
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
478
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
479
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
480 # check early stop condition
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
481 if divergence_flag==max_div_count:
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
482 minibatch_index=nb_max_exemples
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
483 print 'we have diverged, early stopping kicks in'
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
484 break
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
485
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
486 #check if we have seen enough exemples
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
487 #force one epoch at least
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
488 if epoch>0 and minibatch_index*batch_size>nb_max_exemples:
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
489 break
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
490
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
491
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
492
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
493
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
494
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
495 time_n= time_n + batch_size
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
496 minibatch_index = minibatch_index + 1
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
497
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
498 # we have finished looping through the training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
499 epoch = epoch+1
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
500 end_time = time.clock()
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
501
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
502 print(('Optimization complete. Best validation score of %f %% '
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
503 'obtained at iteration %i, with test performance %f %%') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
504 (best_validation_loss * 100., best_iter, test_score*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
505 print ('The code ran for %f minutes' % ((end_time-start_time)/60.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
506 print minibatch_index
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
507 sys.stdout.flush()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
508
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
509 #save the model and the weights
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
510 numpy.savez('model.npy', config=configuration, W1=classifier.W1.value,W2=classifier.W2.value, b1=classifier.b1.value,b2=classifier.b2.value)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
511 numpy.savez('results.npy',config=configuration,total_train_error_list=total_train_error_list,total_validation_error_list=total_validation_error_list,\
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
512 learning_rate_list=learning_rate_list, divergence_flag_list=divergence_flag_list)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
513
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
514 return (best_training_error*100.0,best_validation_loss * 100.,test_score*100.,best_iter*batch_size,(end_time-start_time)/60)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
515
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
516
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
517 if __name__ == '__main__':
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
518 mlp_full_mnist()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
519
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
520 def jobman_mlp_full_nist(state,channel):
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
521 (train_error,validation_error,test_error,nb_exemples,time)=mlp_full_nist(learning_rate=state.learning_rate,\
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
522 nb_max_exemples=state.nb_max_exemples,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
523 nb_hidden=state.nb_hidden,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
524 adaptive_lr=state.adaptive_lr,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
525 tau=state.tau,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
526 verbose = state.verbose,\
324
1763c64030d1 fixed bug in jobman interface
xaviermuller
parents: 323
diff changeset
527 lr_t2_factor=state.lr_t2_factor,
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
528 data_set=state.data_set,
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
529 channel=channel)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
530 state.train_error=train_error
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
531 state.validation_error=validation_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
532 state.test_error=test_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
533 state.nb_exemples=nb_exemples
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
534 state.time=time
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
535 pylearn.version.record_versions(state,[theano,ift6266,pylearn])
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
536 return channel.COMPLETE
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
537
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
538