annotate baseline/conv_mlp/convolutional_mlp.py @ 206:e12702b88a2d

removed one bug: the complexity value was not sent to the dA class
author youssouf
date Thu, 04 Mar 2010 09:43:23 -0500
parents d37c944133c3
children 168aae8a6419
rev   line source
146
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
1 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
2 This tutorial introduces the LeNet5 neural network architecture using Theano. LeNet5 is a
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
3 convolutional neural network, good for classifying images. This tutorial shows how to build the
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
4 architecture, and comes with all the hyper-parameters you need to reproduce the paper's MNIST
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
5 results.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
6
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
7 The best results are obtained after X iterations of the main program loop, which takes ***
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
8 minutes on my workstation (an Intel Core i7, circa July 2009), and *** minutes on my GPU (an
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
9 NVIDIA GTX 285 graphics processor).
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
10
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
11 This implementation simplifies the model in the following ways:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
12
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
13 - LeNetConvPool doesn't implement location-specific gain and bias parameters
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
14 - LeNetConvPool doesn't implement pooling by average, it implements pooling by max.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
15 - Digit classification is implemented with a logistic regression rather than an RBF network
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
16 - LeNet5 was not fully-connected convolutions at second layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
17
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
18 References:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
19 - Y. LeCun, L. Bottou, Y. Bengio and P. Haffner: Gradient-Based Learning Applied to Document
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
20 Recognition, Proceedings of the IEEE, 86(11):2278-2324, November 1998.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
21 http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
22 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
23
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
24 import numpy, theano, cPickle, gzip, time
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
25 import theano.tensor as T
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
26 import theano.sandbox.softsign
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
27 import pylearn.datasets.MNIST
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
28 from pylearn.io import filetensor as ft
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
29 from theano.sandbox import conv, downsample
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
30
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
31 class LeNetConvPoolLayer(object):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
32
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
33 def __init__(self, rng, input, filter_shape, image_shape, poolsize=(2,2)):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
34 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
35 Allocate a LeNetConvPoolLayer with shared variable internal parameters.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
36 :type rng: numpy.random.RandomState
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
37 :param rng: a random number generator used to initialize weights
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
38 :type input: theano.tensor.dtensor4
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
39 :param input: symbolic image tensor, of shape image_shape
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
40 :type filter_shape: tuple or list of length 4
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
41 :param filter_shape: (number of filters, num input feature maps,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
42 filter height,filter width)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
43 :type image_shape: tuple or list of length 4
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
44 :param image_shape: (batch size, num input feature maps,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
45 image height, image width)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
46 :type poolsize: tuple or list of length 2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
47 :param poolsize: the downsampling (pooling) factor (#rows,#cols)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
48 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
49 assert image_shape[1]==filter_shape[1]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
50 self.input = input
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
51
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
52 # initialize weight values: the fan-in of each hidden neuron is
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
53 # restricted by the size of the receptive fields.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
54 fan_in = numpy.prod(filter_shape[1:])
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
55 W_values = numpy.asarray( rng.uniform( \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
56 low = -numpy.sqrt(3./fan_in), \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
57 high = numpy.sqrt(3./fan_in), \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
58 size = filter_shape), dtype = theano.config.floatX)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
59 self.W = theano.shared(value = W_values)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
60
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
61 # the bias is a 1D tensor -- one bias per output feature map
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
62 b_values = numpy.zeros((filter_shape[0],), dtype= theano.config.floatX)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
63 self.b = theano.shared(value= b_values)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
64
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
65 # convolve input feature maps with filters
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
66 conv_out = conv.conv2d(input, self.W,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
67 filter_shape=filter_shape, image_shape=image_shape)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
68
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
69 # downsample each feature map individually, using maxpooling
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
70 pooled_out = downsample.max_pool2D(conv_out, poolsize, ignore_border=True)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
71
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
72 # add the bias term. Since the bias is a vector (1D array), we first
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
73 # reshape it to a tensor of shape (1,n_filters,1,1). Each bias will thus
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
74 # be broadcasted across mini-batches and feature map width & height
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
75 self.output = T.tanh(pooled_out + self.b.dimshuffle('x', 0, 'x', 'x'))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
76
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
77 # store parameters of this layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
78 self.params = [self.W, self.b]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
79
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
80
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
81 class SigmoidalLayer(object):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
82 def __init__(self, rng, input, n_in, n_out):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
83 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
84 Typical hidden layer of a MLP: units are fully-connected and have
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
85 sigmoidal activation function. Weight matrix W is of shape (n_in,n_out)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
86 and the bias vector b is of shape (n_out,).
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
87
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
88 Hidden unit activation is given by: sigmoid(dot(input,W) + b)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
89
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
90 :type rng: numpy.random.RandomState
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
91 :param rng: a random number generator used to initialize weights
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
92 :type input: theano.tensor.dmatrix
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
93 :param input: a symbolic tensor of shape (n_examples, n_in)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
94 :type n_in: int
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
95 :param n_in: dimensionality of input
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
96 :type n_out: int
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
97 :param n_out: number of hidden units
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
98 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
99 self.input = input
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
100
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
101 W_values = numpy.asarray( rng.uniform( \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
102 low = -numpy.sqrt(6./(n_in+n_out)), \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
103 high = numpy.sqrt(6./(n_in+n_out)), \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
104 size = (n_in, n_out)), dtype = theano.config.floatX)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
105 self.W = theano.shared(value = W_values)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
106
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
107 b_values = numpy.zeros((n_out,), dtype= theano.config.floatX)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
108 self.b = theano.shared(value= b_values)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
109
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
110 self.output = T.tanh(T.dot(input, self.W) + self.b)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
111 self.params = [self.W, self.b]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
112
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
113
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
114 class LogisticRegression(object):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
115 """Multi-class Logistic Regression Class
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
116
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
117 The logistic regression is fully described by a weight matrix :math:`W`
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
118 and bias vector :math:`b`. Classification is done by projecting data
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
119 points onto a set of hyperplanes, the distance to which is used to
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
120 determine a class membership probability.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
121 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
122
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
123 def __init__(self, input, n_in, n_out):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
124 """ Initialize the parameters of the logistic regression
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
125 :param input: symbolic variable that describes the input of the
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
126 architecture (one minibatch)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
127 :type n_in: int
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
128 :param n_in: number of input units, the dimension of the space in
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
129 which the datapoints lie
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
130 :type n_out: int
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
131 :param n_out: number of output units, the dimension of the space in
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
132 which the labels lie
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
133 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
134
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
135 # initialize with 0 the weights W as a matrix of shape (n_in, n_out)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
136 self.W = theano.shared( value=numpy.zeros((n_in,n_out),
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
137 dtype = theano.config.floatX) )
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
138 # initialize the baises b as a vector of n_out 0s
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
139 self.b = theano.shared( value=numpy.zeros((n_out,),
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
140 dtype = theano.config.floatX) )
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
141 # compute vector of class-membership probabilities in symbolic form
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
142 self.p_y_given_x = T.nnet.softmax(T.dot(input, self.W)+self.b)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
143
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
144 # compute prediction as class whose probability is maximal in
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
145 # symbolic form
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
146 self.y_pred=T.argmax(self.p_y_given_x, axis=1)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
147
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
148 # list of parameters for this layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
149 self.params = [self.W, self.b]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
150
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
151 def negative_log_likelihood(self, y):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
152 """Return the mean of the negative log-likelihood of the prediction
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
153 of this model under a given target distribution.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
154 :param y: corresponds to a vector that gives for each example the
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
155 correct label
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
156 Note: we use the mean instead of the sum so that
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
157 the learning rate is less dependent on the batch size
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
158 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
159 return -T.mean(T.log(self.p_y_given_x)[T.arange(y.shape[0]),y])
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
160
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
161 def errors(self, y):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
162 """Return a float representing the number of errors in the minibatch
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
163 over the total number of examples of the minibatch ; zero one
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
164 loss over the size of the minibatch
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
165 """
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
166 # check if y has same dimension of y_pred
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
167 if y.ndim != self.y_pred.ndim:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
168 raise TypeError('y should have the same shape as self.y_pred',
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
169 ('y', target.type, 'y_pred', self.y_pred.type))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
170
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
171 # check if y is of the correct datatype
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
172 if y.dtype.startswith('int'):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
173 # the T.neq operator returns a vector of 0s and 1s, where 1
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
174 # represents a mistake in prediction
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
175 return T.mean(T.neq(self.y_pred, y))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
176 else:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
177 raise NotImplementedError()
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
178
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
179
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
180 def load_dataset(fname,batch=20):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
181
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
182 # repertoire qui contient les donnees NIST
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
183 # le repertoire suivant va fonctionner si vous etes connecte sur un ordinateur
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
184 # du reseau DIRO
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
185 datapath = '/data/lisa/data/nist/by_class/'
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
186 # le fichier .ft contient chiffres NIST dans un format efficace. Les chiffres
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
187 # sont stockes dans une matrice de NxD, ou N est le nombre d'images, est D est
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
188 # le nombre de pixels par image (32x32 = 1024). Chaque pixel de l'image est une
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
189 # valeur entre 0 et 255, correspondant a un niveau de gris. Les valeurs sont
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
190 # stockees comme des uint8, donc des bytes.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
191 f = open(datapath+'digits/digits_train_data.ft')
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
192 # Verifier que vous avez assez de memoire pour loader les donnees au complet
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
193 # dans le memoire. Sinon, utilisez ft.arraylike, une classe construite
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
194 # specialement pour des fichiers qu'on ne souhaite pas loader dans RAM.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
195 d = ft.read(f)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
196
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
197 # NB: N'oubliez pas de diviser les valeurs des pixels par 255. si jamais vous
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
198 # utilisez les donnees commes entrees dans un reseaux de neurones et que vous
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
199 # voulez des entres entre 0 et 1.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
200 # digits_train_data.ft contient les images, digits_train_labels.ft contient les
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
201 # etiquettes
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
202 f = open(datapath+'digits/digits_train_labels.ft')
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
203 labels = ft.read(f)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
204
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
205
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
206 # Load the dataset
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
207 #f = gzip.open(fname,'rb')
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
208 #train_set, valid_set, test_set = cPickle.load(f)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
209 #f.close()
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
210
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
211 # make minibatches of size 20
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
212 batch_size = batch # sized of the minibatch
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
213
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
214 # Dealing with the training set
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
215 # get the list of training images (x) and their labels (y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
216 (train_set_x, train_set_y) = (d[:4000,:],labels[:4000])
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
217 # initialize the list of training minibatches with empty list
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
218 train_batches = []
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
219 for i in xrange(0, len(train_set_x), batch_size):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
220 # add to the list of minibatches the minibatch starting at
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
221 # position i, ending at position i+batch_size
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
222 # a minibatch is a pair ; the first element of the pair is a list
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
223 # of datapoints, the second element is the list of corresponding
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
224 # labels
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
225 train_batches = train_batches + \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
226 [(train_set_x[i:i+batch_size], train_set_y[i:i+batch_size])]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
227
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
228 #print train_batches[500]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
229
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
230 # Dealing with the validation set
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
231 (valid_set_x, valid_set_y) = (d[4000:5000,:],labels[4000:5000])
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
232 # initialize the list of validation minibatches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
233 valid_batches = []
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
234 for i in xrange(0, len(valid_set_x), batch_size):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
235 valid_batches = valid_batches + \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
236 [(valid_set_x[i:i+batch_size], valid_set_y[i:i+batch_size])]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
237
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
238 # Dealing with the testing set
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
239 (test_set_x, test_set_y) = (d[5000:6000,:],labels[5000:6000])
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
240 # initialize the list of testing minibatches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
241 test_batches = []
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
242 for i in xrange(0, len(test_set_x), batch_size):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
243 test_batches = test_batches + \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
244 [(test_set_x[i:i+batch_size], test_set_y[i:i+batch_size])]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
245
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
246 return train_batches, valid_batches, test_batches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
247
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
248
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
249 def evaluate_lenet5(learning_rate=0.1, n_iter=1, batch_size=20, n_kern0=20,n_kern1=50,filter_shape=5,n_layer=3, dataset='mnist.pkl.gz'):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
250 rng = numpy.random.RandomState(23455)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
251
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
252 print 'Before load dataset'
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
253 train_batches, valid_batches, test_batches = load_dataset(dataset,batch_size)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
254 print 'After load dataset'
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
255
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
256 ishape = (32,32) # this is the size of NIST images
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
257 n_kern2=80
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
258
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
259 # allocate symbolic variables for the data
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
260 x = T.matrix('x') # rasterized images
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
261 y = T.lvector() # the labels are presented as 1D vector of [long int] labels
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
262
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
263
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
264 ######################
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
265 # BUILD ACTUAL MODEL #
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
266 ######################
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
267
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
268 # Reshape matrix of rasterized images of shape (batch_size,28*28)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
269 # to a 4D tensor, compatible with our LeNetConvPoolLayer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
270 layer0_input = x.reshape((batch_size,1,32,32))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
271
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
272 # Construct the first convolutional pooling layer:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
273 # filtering reduces the image size to (32-5+1,32-5+1)=(28,28)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
274 # maxpooling reduces this further to (28/2,28/2) = (14,14)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
275 # 4D output tensor is thus of shape (20,20,14,14)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
276 layer0 = LeNetConvPoolLayer(rng, input=layer0_input,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
277 image_shape=(batch_size,1,32,32),
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
278 filter_shape=(n_kern0,1,filter_shape,filter_shape), poolsize=(2,2))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
279
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
280 if(n_layer>2):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
281
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
282 # Construct the second convolutional pooling layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
283 # filtering reduces the image size to (14-5+1,14-5+1)=(10,10)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
284 # maxpooling reduces this further to (10/2,10/2) = (5,5)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
285 # 4D output tensor is thus of shape (20,50,5,5)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
286 fshape=(32-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
287 layer1 = LeNetConvPoolLayer(rng, input=layer0.output,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
288 image_shape=(batch_size,n_kern0,fshape,fshape),
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
289 filter_shape=(n_kern1,n_kern0,filter_shape,filter_shape), poolsize=(2,2))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
290
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
291 else:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
292
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
293 fshape=(32-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
294 layer1_input = layer0.output.flatten(2)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
295 # construct a fully-connected sigmoidal layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
296 layer1 = SigmoidalLayer(rng, input=layer1_input,n_in=n_kern0*fshape*fshape, n_out=500)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
297
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
298 layer2 = LogisticRegression(input=layer1.output, n_in=500, n_out=10)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
299 cost = layer2.negative_log_likelihood(y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
300 test_model = theano.function([x,y], layer2.errors(y))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
301 params = layer2.params+ layer1.params + layer0.params
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
302
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
303
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
304 if(n_layer>3):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
305
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
306 fshape=(32-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
307 fshape2=(fshape-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
308 fshape3=(fshape2-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
309 layer2 = LeNetConvPoolLayer(rng, input=layer1.output,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
310 image_shape=(batch_size,n_kern1,fshape2,fshape2),
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
311 filter_shape=(n_kern2,n_kern1,filter_shape,filter_shape), poolsize=(2,2))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
312
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
313 layer3_input = layer2.output.flatten(2)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
314
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
315 layer3 = SigmoidalLayer(rng, input=layer3_input,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
316 n_in=n_kern2*fshape3*fshape3, n_out=500)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
317
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
318
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
319 layer4 = LogisticRegression(input=layer3.output, n_in=500, n_out=10)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
320
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
321 cost = layer4.negative_log_likelihood(y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
322
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
323 test_model = theano.function([x,y], layer4.errors(y))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
324
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
325 params = layer4.params+ layer3.params+ layer2.params+ layer1.params + layer0.params
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
326
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
327
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
328 elif(n_layer>2):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
329
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
330 fshape=(32-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
331 fshape2=(fshape-filter_shape+1)/2
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
332
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
333 # the SigmoidalLayer being fully-connected, it operates on 2D matrices of
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
334 # shape (batch_size,num_pixels) (i.e matrix of rasterized images).
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
335 # This will generate a matrix of shape (20,32*4*4) = (20,512)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
336 layer2_input = layer1.output.flatten(2)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
337
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
338 # construct a fully-connected sigmoidal layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
339 layer2 = SigmoidalLayer(rng, input=layer2_input,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
340 n_in=n_kern1*fshape2*fshape2, n_out=500)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
341
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
342
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
343 # classify the values of the fully-connected sigmoidal layer
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
344 layer3 = LogisticRegression(input=layer2.output, n_in=500, n_out=10)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
345
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
346 # the cost we minimize during training is the NLL of the model
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
347 cost = layer3.negative_log_likelihood(y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
348
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
349 # create a function to compute the mistakes that are made by the model
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
350 test_model = theano.function([x,y], layer3.errors(y))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
351
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
352 # create a list of all model parameters to be fit by gradient descent
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
353 params = layer3.params+ layer2.params+ layer1.params + layer0.params
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
354
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
355
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
356
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
357
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
358
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
359 # create a list of gradients for all model parameters
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
360 grads = T.grad(cost, params)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
361
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
362 # train_model is a function that updates the model parameters by SGD
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
363 # Since this model has many parameters, it would be tedious to manually
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
364 # create an update rule for each model parameter. We thus create the updates
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
365 # dictionary by automatically looping over all (params[i],grads[i]) pairs.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
366 updates = {}
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
367 for param_i, grad_i in zip(params, grads):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
368 updates[param_i] = param_i - learning_rate * grad_i
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
369 train_model = theano.function([x, y], cost, updates=updates)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
370
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
371
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
372 ###############
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
373 # TRAIN MODEL #
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
374 ###############
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
375
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
376 n_minibatches = len(train_batches)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
377
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
378 # early-stopping parameters
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
379 patience = 10000 # look as this many examples regardless
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
380 patience_increase = 2 # wait this much longer when a new best is
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
381 # found
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
382 improvement_threshold = 0.995 # a relative improvement of this much is
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
383 # considered significant
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
384 validation_frequency = n_minibatches # go through this many
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
385 # minibatche before checking the network
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
386 # on the validation set; in this case we
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
387 # check every epoch
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
388
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
389 best_params = None
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
390 best_validation_loss = float('inf')
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
391 best_iter = 0
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
392 test_score = 0.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
393 start_time = time.clock()
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
394
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
395 # have a maximum of `n_iter` iterations through the entire dataset
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
396 for iter in xrange(n_iter * n_minibatches):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
397
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
398 # get epoch and minibatch index
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
399 epoch = iter / n_minibatches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
400 minibatch_index = iter % n_minibatches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
401
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
402 # get the minibatches corresponding to `iter` modulo
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
403 # `len(train_batches)`
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
404 x,y = train_batches[ minibatch_index ]
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
405
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
406 if iter %100 == 0:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
407 print 'training @ iter = ', iter
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
408 cost_ij = train_model(x,y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
409
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
410 if (iter+1) % validation_frequency == 0:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
411
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
412 # compute zero-one loss on validation set
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
413 this_validation_loss = 0.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
414 for x,y in valid_batches:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
415 # sum up the errors for each minibatch
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
416 this_validation_loss += test_model(x,y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
417
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
418 # get the average by dividing with the number of minibatches
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
419 this_validation_loss /= len(valid_batches)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
420 print('epoch %i, minibatch %i/%i, validation error %f %%' % \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
421 (epoch, minibatch_index+1, n_minibatches, \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
422 this_validation_loss*100.))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
423
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
424
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
425 # if we got the best validation score until now
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
426 if this_validation_loss < best_validation_loss:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
427
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
428 #improve patience if loss improvement is good enough
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
429 if this_validation_loss < best_validation_loss * \
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
430 improvement_threshold :
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
431 patience = max(patience, iter * patience_increase)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
432
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
433 # save best validation score and iteration number
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
434 best_validation_loss = this_validation_loss
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
435 best_iter = iter
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
436
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
437 # test it on the test set
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
438 test_score = 0.
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
439 for x,y in test_batches:
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
440 test_score += test_model(x,y)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
441 test_score /= len(test_batches)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
442 print((' epoch %i, minibatch %i/%i, test error of best '
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
443 'model %f %%') %
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
444 (epoch, minibatch_index+1, n_minibatches,
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
445 test_score*100.))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
446
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
447 if patience <= iter :
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
448 break
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
449
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
450 end_time = time.clock()
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
451 print('Optimization complete.')
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
452 print('Best validation score of %f %% obtained at iteration %i,'\
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
453 'with test performance %f %%' %
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
454 (best_validation_loss * 100., best_iter, test_score*100.))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
455 print('The code ran for %f minutes' % ((end_time-start_time)/60.))
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
456
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
457 return (best_validation_loss * 100., test_score*100., (end_time-start_time)/60., best_iter)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
458
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
459 if __name__ == '__main__':
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
460 evaluate_lenet5()
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
461
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
462 def experiment(state, channel):
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
463 print 'start experiment'
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
464 (best_validation_loss, test_score, minutes_trained, iter) = evaluate_lenet5(state.learning_rate, state.n_iter, state.batch_size, state.n_kern0, state.n_kern1, state.filter_shape, state.n_layer)
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
465 print 'end experiment'
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
466
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
467 state.best_validation_loss = best_validation_loss
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
468 state.test_score = test_score
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
469 state.minutes_trained = minutes_trained
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
470 state.iter = iter
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
471
33038ab4e799 Reseau a convolution
Jeremy Eustache <jeremy.eustache@voila.fr>
parents:
diff changeset
472 return channel.COMPLETE