Mercurial > ift6266

annotate baseline_algorithms/mlp/mlp_nist.py @ 145:8ceaaf812891

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
changed adaptive lr flag from bool to int for jobman issues
author XavierMuller
date Tue, 23 Feb 2010 18:23:47 -0500
parents f341a4efb44a
children
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
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
26 import pdb
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
27 import numpy
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
28 import pylab
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
29 import theano
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
30 import theano.tensor as T
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
31 import time
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
32 import theano.tensor.nnet
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
33 import pylearn
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
34 from pylearn.io import filetensor as ft
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
35
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
36 data_path = '/data/lisa/data/nist/by_class/'
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
37
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
38 class MLP(object):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
39 """Multi-Layer Perceptron Class
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
40
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
41 A multilayer perceptron is a feedforward artificial neural network model
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
42 that has one layer or more of hidden units and nonlinear activations.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
43 Intermidiate layers usually have as activation function thanh or the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
44 sigmoid function while the top layer is a softamx layer.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
45 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
46
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
47
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
48
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
49 def __init__(self, input, n_in, n_hidden, n_out,learning_rate):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
50 """Initialize the parameters for the multilayer perceptron
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
51
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
52 :param input: symbolic variable that describes the input of the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
53 architecture (one minibatch)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
54
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
55 :param n_in: number of input units, the dimension of the space in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
56 which the datapoints lie
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
57
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
58 :param n_hidden: number of hidden units
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
59
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
60 :param n_out: number of output units, the dimension of the space in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
61 which the labels lie
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
62
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
63 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
64
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
65 # initialize the parameters theta = (W1,b1,W2,b2) ; note that this
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
66 # example contains only one hidden layer, but one can have as many
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
67 # layers as he/she wishes, making the network deeper. The only
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
68 # problem making the network deep this way is during learning,
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
69 # backpropagation being unable to move the network from the starting
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
70 # point towards; this is where pre-training helps, giving a good
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
71 # starting point for backpropagation, but more about this in the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
72 # other tutorials
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
73
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
74 # `W1` is initialized with `W1_values` which is uniformely sampled
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
75 # from -6./sqrt(n_in+n_hidden) and 6./sqrt(n_in+n_hidden)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
76 # the output of uniform if converted using asarray to dtype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
77 # theano.config.floatX so that the code is runable on GPU
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
78 W1_values = numpy.asarray( numpy.random.uniform( \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
79 low = -numpy.sqrt(6./(n_in+n_hidden)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
80 high = numpy.sqrt(6./(n_in+n_hidden)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
81 size = (n_in, n_hidden)), dtype = theano.config.floatX)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
82 # `W2` is initialized with `W2_values` which is uniformely sampled
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
83 # from -6./sqrt(n_hidden+n_out) and 6./sqrt(n_hidden+n_out)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
84 # the output of uniform if converted using asarray to dtype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
85 # theano.config.floatX so that the code is runable on GPU
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
86 W2_values = numpy.asarray( numpy.random.uniform(
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
87 low = -numpy.sqrt(6./(n_hidden+n_out)), \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
88 high= numpy.sqrt(6./(n_hidden+n_out)),\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
89 size= (n_hidden, n_out)), dtype = theano.config.floatX)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
90
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
91 self.W1 = theano.shared( value = W1_values )
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
92 self.b1 = theano.shared( value = numpy.zeros((n_hidden,),
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
93 dtype= theano.config.floatX))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
94 self.W2 = theano.shared( value = W2_values )
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
95 self.b2 = theano.shared( value = numpy.zeros((n_out,),
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
96 dtype= theano.config.floatX))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
97
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
98 #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
99 #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
100 lr_value=learning_rate
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
101 self.lr=theano.shared(value=lr_value)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
102 # symbolic expression computing the values of the hidden layer
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
103 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
104
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
105
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
106
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
107 # symbolic expression computing the values of the top layer
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
108 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
109
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
110 # compute prediction as class whose probability is maximal in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
111 # symbolic form
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
112 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
113 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
114
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
115
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
116
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
117
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
118 # L1 norm ; one regularization option is to enforce L1 norm to
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
119 # be small
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
120 self.L1 = abs(self.W1).sum() + abs(self.W2).sum()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
121
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
122 # square of L2 norm ; one regularization option is to enforce
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
123 # square of L2 norm to be small
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
124 self.L2_sqr = (self.W1**2).sum() + (self.W2**2).sum()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
125
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
126
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
127
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
128 def negative_log_likelihood(self, y):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
129 """Return the mean of the negative log-likelihood of the prediction
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
130 of this model under a given target distribution.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
131
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
132 .. math::
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
133
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
134 \frac{1}{|\mathcal{D}|}\mathcal{L} (\theta=\{W,b\}, \mathcal{D}) =
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
135 \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
136 \ell (\theta=\{W,b\}, \mathcal{D})
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
137
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
138
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
139 :param y: corresponds to a vector that gives for each example the
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
140 :correct label
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
141 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
142 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
143
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
144
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
145
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
146
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
147 def errors(self, y):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
148 """Return a float representing the number of errors in the minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
149 over the total number of examples of the minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
150 """
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
151
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
152 # check if y has same dimension of y_pred
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
153 if y.ndim != self.y_pred.ndim:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
154 raise TypeError('y should have the same shape as self.y_pred',
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
155 ('y', target.type, 'y_pred', self.y_pred.type))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
156 # check if y is of the correct datatype
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
157 if y.dtype.startswith('int'):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
158 # the T.neq operator returns a vector of 0s and 1s, where 1
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
159 # represents a mistake in prediction
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
160 return T.mean(T.neq(self.y_pred, y))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
161 else:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
162 raise NotImplementedError()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
163
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
164
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
165 def mlp_full_nist( verbose = False,\
145
8ceaaf812891 changed adaptive lr flag from bool to int for jobman issues
XavierMuller
parents: 143
diff changeset
166 adaptive_lr = 0,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
167 train_data = 'all/all_train_data.ft',\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
168 train_labels = 'all/all_train_labels.ft',\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
169 test_data = 'all/all_test_data.ft',\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
170 test_labels = 'all/all_test_labels.ft',\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
171 learning_rate=0.01,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
172 L1_reg = 0.00,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
173 L2_reg = 0.0001,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
174 nb_max_exemples=1000000,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
175 batch_size=20,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
176 nb_hidden = 500,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
177 nb_targets = 62):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
178
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
179
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
180 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
181
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
182 total_validation_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
183 total_train_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
184 learning_rate_list=[]
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
185 best_training_error=float('inf');
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
186
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
187
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
188
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
189 f = open(data_path+train_data)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
190 g= open(data_path+train_labels)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
191 h = open(data_path+test_data)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
192 i= open(data_path+test_labels)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
193
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
194 raw_train_data = ft.read(f)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
195 raw_train_labels = ft.read(g)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
196 raw_test_data = ft.read(h)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
197 raw_test_labels = ft.read(i)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
198
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
199 f.close()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
200 g.close()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
201 i.close()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
202 h.close()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
203 #create a validation set the same size as the test size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
204 #use the end of the training array for this purpose
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
205 #discard the last remaining so we get a %batch_size number
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
206 test_size=len(raw_test_labels)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
207 test_size = int(test_size/batch_size)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
208 test_size*=batch_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
209 train_size = len(raw_train_data)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
210 train_size = int(train_size/batch_size)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
211 train_size*=batch_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
212 validation_size =test_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
213 offset = train_size-test_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
214 if verbose == True:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
215 print 'train size = %d' %train_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
216 print 'test size = %d' %test_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
217 print 'valid size = %d' %validation_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
218 print 'offset = %d' %offset
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
219
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
220
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
221 train_set = (raw_train_data,raw_train_labels)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
222 train_batches = []
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
223 for i in xrange(0, train_size-test_size, batch_size):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
224 train_batches = train_batches + \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
225 [(raw_train_data[i:i+batch_size], raw_train_labels[i:i+batch_size])]
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
226
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
227 test_batches = []
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
228 for i in xrange(0, test_size, batch_size):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
229 test_batches = test_batches + \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
230 [(raw_test_data[i:i+batch_size], raw_test_labels[i:i+batch_size])]
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
231
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
232 validation_batches = []
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
233 for i in xrange(0, test_size, batch_size):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
234 validation_batches = validation_batches + \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
235 [(raw_train_data[offset+i:offset+i+batch_size], raw_train_labels[offset+i:offset+i+batch_size])]
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
236
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
237
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
238 ishape = (32,32) # this is the size of NIST images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
239
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
240 # allocate symbolic variables for the data
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
241 x = T.fmatrix() # the data is presented as rasterized images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
242 y = T.lvector() # the labels are presented as 1D vector of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
243 # [long int] labels
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
244
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
245 if verbose==True:
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
246 print 'finished parsing the data'
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
247 # construct the logistic regression class
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
248 classifier = MLP( input=x.reshape((batch_size,32*32)),\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
249 n_in=32*32,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
250 n_hidden=nb_hidden,\
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
251 n_out=nb_targets,
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
252 learning_rate=learning_rate)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
253
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
254
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
255
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
256
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
257 # the cost we minimize during training is the negative log likelihood of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
258 # the model plus the regularization terms (L1 and L2); cost is expressed
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
259 # here symbolically
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
260 cost = classifier.negative_log_likelihood(y) \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
261 + L1_reg * classifier.L1 \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
262 + L2_reg * classifier.L2_sqr
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
263
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
264 # compiling a theano function that computes the mistakes that are made by
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
265 # the model on a minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
266 test_model = theano.function([x,y], classifier.errors(y))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
267
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
268 # compute the gradient of cost with respect to theta = (W1, b1, W2, b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
269 g_W1 = T.grad(cost, classifier.W1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
270 g_b1 = T.grad(cost, classifier.b1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
271 g_W2 = T.grad(cost, classifier.W2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
272 g_b2 = T.grad(cost, classifier.b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
273
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
274 # specify how to update the parameters of the model as a dictionary
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
275 updates = \
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
276 { 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
277 , 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
278 , 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
279 , classifier.b2: classifier.b2 - classifier.lr*g_b2 }
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
280
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
281 # compiling a theano function `train_model` that returns the cost, but in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
282 # the same time updates the parameter of the model based on the rules
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
283 # defined in `updates`
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
284 train_model = theano.function([x, y], cost, updates = updates )
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
285 n_minibatches = len(train_batches)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
286
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
287
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
288
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
289
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
290
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
291
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
292 #conditions for stopping the adaptation:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
293 #1) we have reached nb_max_exemples (this is rounded up to be a multiple of the train size)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
294 #2) validation error is going up twice in a row(probable overfitting)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
295
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
296 # This means we no longer stop on slow convergence as low learning rates stopped
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
297 # too fast.
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
298
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
299 # no longer relevant
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
300 patience =nb_max_exemples/batch_size
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
301 patience_increase = 2 # wait this much longer when a new best is
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
302 # found
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
303 improvement_threshold = 0.995 # a relative improvement of this much is
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
304 # considered significant
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
305 validation_frequency = n_minibatches/4
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
306
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
307
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
308
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
309
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
310 best_params = None
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
311 best_validation_loss = float('inf')
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
312 best_iter = 0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
313 test_score = 0.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
314 start_time = time.clock()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
315 n_iter = nb_max_exemples/batch_size # nb of max times we are allowed to run through all exemples
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
316 n_iter = n_iter/n_minibatches + 1 #round up
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
317 n_iter=max(1,n_iter) # run at least once on short debug call
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
318
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
319
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
320 if verbose == True:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
321 print 'looping at most %d times through the data set' %n_iter
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
322 for iter in xrange(n_iter* n_minibatches):
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
323
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
324 # get epoch and minibatch index
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
325 epoch = iter / n_minibatches
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
326 minibatch_index = iter % n_minibatches
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
327
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
328
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
329
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
330 # get the minibatches corresponding to `iter` modulo
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
331 # `len(train_batches)`
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
332 x,y = train_batches[ minibatch_index ]
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
333 # convert to float
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
334 x_float = x/255.0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
335 cost_ij = train_model(x_float,y)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
336
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
337 if (iter+1) % validation_frequency == 0:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
338 # compute zero-one loss on validation set
143
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 this_validation_loss = 0.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
341 for x,y in validation_batches:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
342 # sum up the errors for each minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
343 x_float = x/255.0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
344 this_validation_loss += test_model(x_float,y)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
345 # get the average by dividing with the number of minibatches
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
346 this_validation_loss /= len(validation_batches)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
347 #save the validation loss
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
348 total_validation_error_list.append(this_validation_loss)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
349
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
350 #get the training error rate
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
351 this_train_loss=0
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
352 for x,y in train_batches:
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
353 # sum up the errors for each minibatch
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
354 x_float = x/255.0
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
355 this_train_loss += test_model(x_float,y)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
356 # get the average by dividing with the number of minibatches
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
357 this_train_loss /= len(train_batches)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
358 #save the validation loss
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
359 total_train_error_list.append(this_train_loss)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
360 if(this_train_loss<best_training_error):
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
361 best_training_error=this_train_loss
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
362
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
363 if verbose == True:
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
364 print('epoch %i, minibatch %i/%i, validation error %f, training error %f %%' % \
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
365 (epoch, minibatch_index+1, n_minibatches, \
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
366 this_validation_loss*100.,this_train_loss*100))
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
367
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
368
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
369 #save the learning rate
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
370 learning_rate_list.append(classifier.lr.value)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
371
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
372
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
373 # if we got the best validation score until now
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
374 if this_validation_loss < best_validation_loss:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
375 # save best validation score and iteration number
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
376 best_validation_loss = this_validation_loss
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
377 best_iter = iter
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
378 # reset patience if we are going down again
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
379 # so we continue exploring
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
380 patience=nb_max_exemples/batch_size
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
381 # test it on the test set
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
382 test_score = 0.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
383 for x,y in test_batches:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
384 x_float=x/255.0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
385 test_score += test_model(x_float,y)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
386 test_score /= len(test_batches)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
387 if verbose == True:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
388 print((' epoch %i, minibatch %i/%i, test error of best '
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
389 'model %f %%') %
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
390 (epoch, minibatch_index+1, n_minibatches,
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
391 test_score*100.))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
392
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
393 # if the validation error is going up, we are overfitting (or oscillating)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
394 # stop converging but run at least to next validation
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
395 # to check overfitting or ocsillation
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
396 # the saved weights of the model will be a bit off in that case
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
397 elif this_validation_loss >= best_validation_loss:
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
398 #calculate the test error at this point and exit
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
399 # test it on the test set
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
400 # however, if adaptive_lr is true, try reducing the lr to
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
401 # get us out of an oscilliation
145
8ceaaf812891 changed adaptive lr flag from bool to int for jobman issues
XavierMuller
parents: 143
diff changeset
402 if adaptive_lr==1:
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
403 classifier.lr.value=classifier.lr.value/2.0
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
404
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
405 test_score = 0.
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
406 #cap the patience so we are allowed one more validation error
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
407 #calculation before aborting
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
408 patience = iter+validation_frequency+1
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
409 for x,y in test_batches:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
410 x_float=x/255.0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
411 test_score += test_model(x_float,y)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
412 test_score /= len(test_batches)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
413 if verbose == True:
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
414 print ' validation error is going up, possibly stopping soon'
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
415 print((' epoch %i, minibatch %i/%i, test error of best '
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
416 'model %f %%') %
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
417 (epoch, minibatch_index+1, n_minibatches,
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
418 test_score*100.))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
419
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
420
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
421
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
422
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
423 if iter>patience:
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
424 print 'we have diverged'
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
425 break
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
426
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
427
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
428 end_time = time.clock()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
429 if verbose == True:
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
430 print(('Optimization complete. Best validation score of %f %% '
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
431 'obtained at iteration %i, with test performance %f %%') %
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
432 (best_validation_loss * 100., best_iter, test_score*100.))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
433 print ('The code ran for %f minutes' % ((end_time-start_time)/60.))
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
434 print iter
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
435
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
436 #save the model and the weights
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
437 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
438 numpy.savez('results.npy',config=configuration,total_train_error_list=total_train_error_list,total_validation_error_list=total_validation_error_list,\
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
439 learning_rate_list=learning_rate_list)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
440
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
441 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
442
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
443
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
444 if __name__ == '__main__':
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
445 mlp_full_mnist()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
446
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
447 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
448 (train_error,validation_error,test_error,nb_exemples,time)=mlp_full_nist(learning_rate=state.learning_rate,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
449 nb_max_exemples=state.nb_max_exemples,\
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
450 nb_hidden=state.nb_hidden,\
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
451 adaptive_lr=state.adaptive_lr)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
452 state.train_error=train_error
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
453 state.validation_error=validation_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
454 state.test_error=test_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
455 state.nb_exemples=nb_exemples
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
456 state.time=time
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
457 return channel.COMPLETE
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
458
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
459