Mercurial > ift6266

annotate baseline/mlp/mlp_nist.py @ 417:0282882aa91f

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Completed the pinch transformation text
author fsavard
date Fri, 30 Apr 2010 09:25:20 -0400
parents 3dba84c0fbc1
children 868f82777839
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
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
172
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
173
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
174
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
175
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
176 # 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
177 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
178 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
179 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
180 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
181 b2=model['b2']
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
182
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
183 total_error_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
184 total_exemple_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
185
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
186 nb_error_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
187 nb_exemple_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
188
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
189 char_error_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
190 char_exemple_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
191
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
192 min_error_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
193 min_exemple_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
194
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
195 maj_error_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
196 maj_exemple_count=0.0
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
197
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
198 vtotal_error_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
199 vtotal_exemple_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
200
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
201 vnb_error_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
202 vnb_exemple_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
203
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
204 vchar_error_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
205 vchar_exemple_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
206
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
207 vmin_error_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
208 vmin_exemple_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
209
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
210 vmaj_error_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
211 vmaj_exemple_count=0.0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
212
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
213
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
214
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
215 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
216 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
217 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
218 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
219
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
220
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
221
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
222 #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
223 #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
224 #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
225 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
226 temp=0
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
227 for xt,yt in dataset.test(1):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
228
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
229 total_exemple_count = total_exemple_count +1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
230 #get activation for layer 1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
231 a0=numpy.dot(numpy.transpose(W1),numpy.transpose(xt[0])) + b1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
232 #add non linear function to layer 1 activation
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
233 a0_out=numpy.tanh(a0)
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
234
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
235 #get activation for output layer
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
236 a1= numpy.dot(numpy.transpose(W2),a0_out) + b2
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
237 #add non linear function for output activation (softmax)
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
238 a1_exp = numpy.exp(a1)
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
239 sum_a1=numpy.sum(a1_exp)
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
240 a1_out=a1_exp/sum_a1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
241
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
242 predicted_class=numpy.argmax(a1_out)
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
243 wanted_class=yt[0]
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
244 if(predicted_class!=wanted_class):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
245 total_error_count = total_error_count +1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
246
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
247 #treat digit error
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
248 if(wanted_class<10):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
249 nb_exemple_count=nb_exemple_count + 1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
250 predicted_class=numpy.argmax(a1_out[0:10])
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
251 if(predicted_class!=wanted_class):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
252 nb_error_count = nb_error_count +1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
253
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
254 if(wanted_class>9):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
255 char_exemple_count=char_exemple_count + 1
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
256 predicted_class=numpy.argmax(a1_out[10:62])+10
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
257 if((predicted_class!=wanted_class) and ((predicted_class+26)!=wanted_class) and ((predicted_class-26)!=wanted_class)):
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
258 char_error_count = char_error_count +1
405
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
259
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
260 #minuscule
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
261 if(wanted_class>9 and wanted_class<36):
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
262 maj_exemple_count=maj_exemple_count + 1
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
263 predicted_class=numpy.argmax(a1_out[10:35])+10
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
264 if(predicted_class!=wanted_class):
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
265 maj_error_count = maj_error_count +1
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
266 #majuscule
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
267 if(wanted_class>35):
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
268 min_exemple_count=min_exemple_count + 1
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
269 predicted_class=numpy.argmax(a1_out[36:62])+36
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
270 if(predicted_class!=wanted_class):
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
271 min_error_count = min_error_count +1
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
272
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
273
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
274
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
275 vtest_score=0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
276 vtemp=0
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
277 for xt,yt in dataset.valid(1):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
278
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
279 vtotal_exemple_count = vtotal_exemple_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
280 #get activation for layer 1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
281 a0=numpy.dot(numpy.transpose(W1),numpy.transpose(xt[0])) + b1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
282 #add non linear function to layer 1 activation
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
283 a0_out=numpy.tanh(a0)
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
284
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
285 #get activation for output layer
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
286 a1= numpy.dot(numpy.transpose(W2),a0_out) + b2
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
287 #add non linear function for output activation (softmax)
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
288 a1_exp = numpy.exp(a1)
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
289 sum_a1=numpy.sum(a1_exp)
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
290 a1_out=a1_exp/sum_a1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
291
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
292 predicted_class=numpy.argmax(a1_out)
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
293 wanted_class=yt[0]
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
294 if(predicted_class!=wanted_class):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
295 vtotal_error_count = vtotal_error_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
296
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
297 #treat digit error
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
298 if(wanted_class<10):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
299 vnb_exemple_count=vnb_exemple_count + 1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
300 predicted_class=numpy.argmax(a1_out[0:10])
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
301 if(predicted_class!=wanted_class):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
302 vnb_error_count = vnb_error_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
303
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
304 if(wanted_class>9):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
305 vchar_exemple_count=vchar_exemple_count + 1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
306 predicted_class=numpy.argmax(a1_out[10:62])+10
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
307 if((predicted_class!=wanted_class) and ((predicted_class+26)!=wanted_class) and ((predicted_class-26)!=wanted_class)):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
308 vchar_error_count = vchar_error_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
309
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
310 #minuscule
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
311 if(wanted_class>9 and wanted_class<36):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
312 vmaj_exemple_count=vmaj_exemple_count + 1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
313 predicted_class=numpy.argmax(a1_out[10:35])+10
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
314 if(predicted_class!=wanted_class):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
315 vmaj_error_count = vmaj_error_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
316 #majuscule
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
317 if(wanted_class>35):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
318 vmin_exemple_count=vmin_exemple_count + 1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
319 predicted_class=numpy.argmax(a1_out[36:62])+36
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
320 if(predicted_class!=wanted_class):
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
321 vmin_error_count = vmin_error_count +1
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
322
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
323
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
324 print (('total error = %f') % ((total_error_count/total_exemple_count)*100.0))
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
325 print (('number error = %f') % ((nb_error_count/nb_exemple_count)*100.0))
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
326 print (('char error = %f') % ((char_error_count/char_exemple_count)*100.0))
405
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
327 print (('min error = %f') % ((min_error_count/min_exemple_count)*100.0))
195f95c3d461 corrected bug where we inverted min and maj
xaviermuller
parents: 404
diff changeset
328 print (('maj error = %f') % ((maj_error_count/maj_exemple_count)*100.0))
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
329
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
330 print (('valid total error = %f') % ((vtotal_error_count/vtotal_exemple_count)*100.0))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
331 print (('valid number error = %f') % ((vnb_error_count/vnb_exemple_count)*100.0))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
332 print (('valid char error = %f') % ((vchar_error_count/vchar_exemple_count)*100.0))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
333 print (('valid min error = %f') % ((vmin_error_count/vmin_exemple_count)*100.0))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
334 print (('valid maj error = %f') % ((vmaj_error_count/vmaj_exemple_count)*100.0))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
335
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
336 print ((' num total = %d,%d') % (total_exemple_count,total_error_count))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
337 print ((' num nb = %d,%d') % (nb_exemple_count,nb_error_count))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
338 print ((' num min = %d,%d') % (min_exemple_count,min_error_count))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
339 print ((' num maj = %d,%d') % (maj_exemple_count,maj_error_count))
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
340 print ((' num char = %d,%d') % (char_exemple_count,char_error_count))
404
1509b9bba4cc added digit/char error
xaviermuller
parents: 378
diff changeset
341 return (total_error_count/total_exemple_count)*100.0
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
342
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
343
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
344
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
345
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
346
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
347
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
348 def mlp_full_nist( verbose = 1,\
145
8ceaaf812891 changed adaptive lr flag from bool to int for jobman issues
XavierMuller
parents: 143
diff changeset
349 adaptive_lr = 0,\
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
350 data_set=0,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
351 learning_rate=0.01,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
352 L1_reg = 0.00,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
353 L2_reg = 0.0001,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
354 nb_max_exemples=1000000,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
355 batch_size=20,\
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
356 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
357 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
358 tau=1e6,\
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
359 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
360 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
361 channel=0):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
362
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
363
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
364 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
365 channel.save()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
366 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
367
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 #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
369 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
370 max_div_count=1000
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
371 optimal_test_error=0
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
372
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
373
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
374 total_validation_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
375 total_train_error_list = []
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
376 learning_rate_list=[]
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
377 best_training_error=float('inf');
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
378 divergence_flag_list=[]
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
379
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
380 if data_set==0:
378
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
381 print 'using nist'
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
382 dataset=datasets.nist_all()
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
383 elif data_set==1:
378
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
384 print 'using p07'
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
385 dataset=datasets.nist_P07()
349
22efb4968054 added pnist support, will check in code for data set iterator later
xaviermuller
parents: 338
diff changeset
386 elif data_set==2:
378
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
387 print 'using pnist'
349
22efb4968054 added pnist support, will check in code for data set iterator later
xaviermuller
parents: 338
diff changeset
388 dataset=datasets.PNIST07()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
389
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
390
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
391
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
392
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
393 ishape = (32,32) # this is the size of NIST images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
394
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
395 # allocate symbolic variables for the data
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
396 x = T.fmatrix() # the data is presented as rasterized images
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
397 y = T.lvector() # the labels are presented as 1D vector of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
398 # [long int] labels
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
399
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
400
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
401 # construct the logistic regression class
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
402 classifier = MLP( input=x,\
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
403 n_in=32*32,\
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
404 n_hidden=nb_hidden,\
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
405 n_out=nb_targets,
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
406 learning_rate=learning_rate)
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
407
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
408
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
409 # 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
410 # 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
411 if init_model!=0:
378
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
412 print 'using old model'
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
413 print init_model
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
414 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
415 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
416 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
417 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
418 classifier.b2.value=old_model['b2']
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
419
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
420
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
421 # the cost we minimize during training is the negative log likelihood of
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
422 # the model plus the regularization terms (L1 and L2); cost is expressed
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
423 # here symbolically
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
424 cost = classifier.negative_log_likelihood(y) \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
425 + L1_reg * classifier.L1 \
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
426 + L2_reg * classifier.L2_sqr
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
427
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
428 # compiling a theano function that computes the mistakes that are made by
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
429 # the model on a minibatch
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
430 test_model = theano.function([x,y], classifier.errors(y))
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
431
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
432 # compute the gradient of cost with respect to theta = (W1, b1, W2, b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
433 g_W1 = T.grad(cost, classifier.W1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
434 g_b1 = T.grad(cost, classifier.b1)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
435 g_W2 = T.grad(cost, classifier.W2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
436 g_b2 = T.grad(cost, classifier.b2)
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
437
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
438 # specify how to update the parameters of the model as a dictionary
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
439 updates = \
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
440 { 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
441 , 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
442 , 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
443 , classifier.b2: classifier.b2 - classifier.lr*g_b2 }
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
444
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
445 # compiling a theano function `train_model` that returns the cost, but in
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
446 # the same time updates the parameter of the model based on the rules
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
447 # defined in `updates`
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
448 train_model = theano.function([x, y], cost, updates = updates )
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
449
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
450
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
451
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
452
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
453
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
454
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
455
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
456
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
457
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
458 #conditions for stopping the adaptation:
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
459 #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
460 #2) validation error is going up twice in a row(probable overfitting)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
461
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
462 # 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
463 # 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
464 # 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
465 # 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
466 # a bit when diverging
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
467
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
468 #approximate number of samples in the nist training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
469 #this is just to have a validation frequency
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
470 #roughly proportionnal to the original nist training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
471 n_minibatches = 650000/batch_size
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
472
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
473
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
474 patience =2*nb_max_exemples/batch_size #in units of minibatch
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
475 validation_frequency = n_minibatches/4
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
476
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
477
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
478
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
479
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
480
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
481 best_validation_loss = float('inf')
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
482 best_iter = 0
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
483 test_score = 0.
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
484 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
485 time_n=0 #in unit of exemples
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
486 minibatch_index=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
487 epoch=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
488 temp=0
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
489 divergence_flag=0
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
490
212
e390b0454515 added classic lr time decay and py code to calculate the error based on a saved model
xaviermuller
parents: 169
diff changeset
491
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
492
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
493
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
494 print 'starting training'
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
495 sys.stdout.flush()
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
496 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
497
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
498 for x, y in dataset.train(batch_size):
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
499
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
500 #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
501 if adaptive_lr==2:
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
502 classifier.lr.value = tau*initial_lr/(tau+time_n)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
503
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
504
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
505 #train model
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
506 cost_ij = train_model(x,y)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
507
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
508 if (minibatch_index) % validation_frequency == 0:
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
509 #save the current learning rate
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
510 learning_rate_list.append(classifier.lr.value)
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
511 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
512
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
513
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
514
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
515 # compute the validation error
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
516 this_validation_loss = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
517 temp=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
518 for xv,yv in dataset.valid(1):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
519 # sum up the errors for each minibatch
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
520 this_validation_loss += test_model(xv,yv)
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
521 temp=temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
522 # get the average by dividing with the number of minibatches
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
523 this_validation_loss /= temp
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
524 #save the validation loss
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
525 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
526
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
527 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
528 (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
529 this_validation_loss*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
530 sys.stdout.flush()
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
531
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
532 #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
533 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
534 learning_rate_list=learning_rate_list, divergence_flag_list=divergence_flag_list)
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
535
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
536 # if we got the best validation score until now
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
537 if this_validation_loss < best_validation_loss:
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
538 # save best validation score and iteration number
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
539 best_validation_loss = this_validation_loss
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
540 best_iter = minibatch_index
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
541 #reset divergence flag
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
542 divergence_flag=0
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
543
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
544 #save the best model. Overwrite the current saved best model so
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
545 #we only keep the best
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
546 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
547 b2=classifier.b2.value, minibatch_index=minibatch_index)
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
548
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
549 # test it on the test set
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
550 test_score = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
551 temp =0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
552 for xt,yt in dataset.test(batch_size):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
553 test_score += test_model(xt,yt)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
554 temp = temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
555 test_score /= temp
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
556
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
557 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
558 'model %f %%') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
559 (epoch, minibatch_index+1,
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
560 test_score*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
561 sys.stdout.flush()
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
562 optimal_test_error=test_score
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
563
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
564 # 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
565 # 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
566 elif this_validation_loss >= best_validation_loss:
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
567
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
568
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
569 # In non-classic learning rate decay, we modify the weight only when
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
570 # validation error is going up
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
571 if adaptive_lr==1:
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
572 classifier.lr.value=classifier.lr.value*lr_t2_factor
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
573
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
574
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
575 #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
576 #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
577 divergence_flag = divergence_flag +1
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
578
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
579
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
580 #calculate the test error at this point and exit
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
581 # test it on the test set
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
582 test_score = 0.
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
583 temp=0
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
584 for xt,yt in dataset.test(batch_size):
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
585 test_score += test_model(xt,yt)
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
586 temp=temp+1
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
587 test_score /= temp
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
588
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
589 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
590 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
591 'model %f %%') %
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
592 (epoch, minibatch_index+1,
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
593 test_score*100.))
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
594 sys.stdout.flush()
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
595
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
596
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
597
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
598 # check early stop condition
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
599 if divergence_flag==max_div_count:
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
600 minibatch_index=nb_max_exemples
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
601 print 'we have diverged, early stopping kicks in'
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
602 break
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
603
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
604 #check if we have seen enough exemples
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
605 #force one epoch at least
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
606 if epoch>0 and minibatch_index*batch_size>nb_max_exemples:
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
607 break
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
608
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
609
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
610
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
611
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
612
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
613 time_n= time_n + batch_size
323
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
614 minibatch_index = minibatch_index + 1
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
615
7a7615f940e8 finished code clean up and testing
xaviermuller
parents: 322
diff changeset
616 # we have finished looping through the training set
322
743907366476 code clean up in progress
xaviermuller
parents: 304
diff changeset
617 epoch = epoch+1
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
618 end_time = time.clock()
355
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
619
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
620 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
621 '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
622 (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
623 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
624 print minibatch_index
76b7182dd32e added support for pnist in iterator. corrected a print bug in mlp
xaviermuller
parents: 349
diff changeset
625 sys.stdout.flush()
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
626
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
627 #save the model and the weights
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
628 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
629 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
630 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
631
414
3dba84c0fbc1 saving test score from best validation score in db now
xaviermuller
parents: 405
diff changeset
632 return (best_training_error*100.0,best_validation_loss * 100.,optimal_test_error*100.,best_iter*batch_size,(end_time-start_time)/60)
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
633
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
634
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
635 if __name__ == '__main__':
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
636 mlp_full_mnist()
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
637
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
638 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
639 (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
640 nb_max_exemples=state.nb_max_exemples,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
641 nb_hidden=state.nb_hidden,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
642 adaptive_lr=state.adaptive_lr,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
643 tau=state.tau,\
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
644 verbose = state.verbose,\
324
1763c64030d1 fixed bug in jobman interface
xaviermuller
parents: 323
diff changeset
645 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
646 data_set=state.data_set,
378
60a4432b8071 added initial model for weights in jobman
xaviermuller
parents: 355
diff changeset
647 init_model=state.init_model,
338
fca22114bb23 added async save, restart from old model and independant error calculation based on Arnaud's iterator
xaviermuller
parents: 324
diff changeset
648 channel=channel)
143
f341a4efb44a added adaptive lr, weight file save, traine error and error curves
XavierMuller
parents: 110
diff changeset
649 state.train_error=train_error
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
650 state.validation_error=validation_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
651 state.test_error=test_error
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
652 state.nb_exemples=nb_exemples
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
653 state.time=time
304
1e4bf5a5b46d added type 2 adaptive learning configurable learning weight + versionning
xaviermuller
parents: 237
diff changeset
654 pylearn.version.record_versions(state,[theano,ift6266,pylearn])
110
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
655 return channel.COMPLETE
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
656
93b4b84d86cf added simple mlp file
XavierMuller
parents:
diff changeset
657