def test_SdA(finetune_lr=0.1, pretraining_epochs=20, ## originally 15
pretrain_lr=0.001, training_epochs=1000,
dataset='cifar-10-batches-py', batch_size=1, mode='train', amount='full'):
"""
Demonstrates how to train and test a stochastic denoising autoencoder.
This is demonstrated on MNIST.
:type learning_rate: float
:param learning_rate: learning rate used in the finetune stage
(factor for the stochastic gradient)
:type pretraining_epochs: int
:param pretraining_epochs: number of epoch to do pretraining
:type pretrain_lr: float
:param pretrain_lr: learning rate to be used during pre-training
:type n_iter: int
:param n_iter: maximal number of iterations ot run the optimizer
:type dataset: string
:param dataset: path the the pickled dataset
"""
datasets = load_data(dataset, mode=mode, amount=amount)
if mode == 'train':
train_set_x, train_set_y = datasets[0]
valid_set_x, valid_set_y = datasets[1]
else:
test_set_x, test_set_y = datasets[0]
# compute number of minibatches for training, validation and testing
if mode == 'train':
n_train_batches = train_set_x.get_value(borrow=True).shape[0]
n_train_batches /= batch_size
else:
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
# numpy random generator
numpy_rng = numpy.random.RandomState(89677)
# allocate symbolic variables for the data
index = T.lscalar() # index to a [mini]batch
x = T.matrix('x') # the data is presented as rasterized images
y = T.ivector('y') # the labels are presented as 1D vector of
print '... building the model'
# construct the stacked denoising autoencoder class
sda = SdA(numpy_rng=numpy_rng, n_ins=32 * 32,
hidden_layers_sizes=[1300, 1300, 1300],
n_outs=10)
## load the saved parameters
if mode == 'test':
learned_params = unpickle('params/SdA.pkl')
print '... getting the pretraining functions'
if mode == 'train':
pretraining_fns = sda.pretraining_functions(train_set_x=train_set_x,
batch_size=batch_size)
#########################
# PRETRAINING THE MODEL #
#########################
if mode == 'train':
print '... pre-training the model'
start_time = time.clock()
## Pre-train layer-wise
corruption_levels = [.1, .2, .3]
for i in xrange(sda.n_layers):
# go through pretraining epochs
for epoch in xrange(pretraining_epochs):
# go through the training set
c = []
for batch_index in xrange(n_train_batches):
c.append(pretraining_fns[i](index=batch_index,
corruption=corruption_levels[i],
lr=pretrain_lr))
print 'Pre-training layer %i, epoch %d / %d, cost ' % (i, epoch + 1, pretraining_epochs),
print numpy.mean(c)
end_time = time.clock()
print >> sys.stderr, ('The pretraining code for file ' +
os.path.split(__file__)[1] +
' ran for %.2fm' % ((end_time - start_time) / 60.))
########################
# FINETUNING THE MODEL #
########################
# get the training, validation and testing function for the model
print '... getting the finetuning functions'
if mode == 'train':
train_fn, validate_model = sda.build_finetune_functions(
datasets=datasets, batch_size=batch_size,
learning_rate=finetune_lr)
print '... finetunning the model'
# early-stopping parameters
if mode == 'train':
patience = 10 * n_train_batches # look as this many examples regardless
patience_increase = 2. # wait this much longer when a new best is
# found
improvement_threshold = 0.995 # a relative improvement of this much is
# considered significant
validation_frequency = min(n_train_batches, patience / 2)
# go through this many
# minibatche before checking the network
# on the validation set; in this case we
# check every epoch
# create a function to get the labels predicted by the model
if mode == 'test':
get_test_labels = theano.function([index], sda.logLayer.y_pred,
givens={
x: test_set_x[index * batch_size: (index + 1) * batch_size],
sda.sigmoid_layers[0].W: learned_params[0],
sda.sigmoid_layers[0].b: learned_params[1],
sda.sigmoid_layers[1].W: learned_params[2],
sda.sigmoid_layers[1].b: learned_params[3],
sda.sigmoid_layers[2].W: learned_params[4],
sda.sigmoid_layers[2].b: learned_params[5],
sda.logLayer.W: learned_params[6],
sda.logLayer.b: learned_params[7]})
best_params = None
best_validation_loss = numpy.inf
test_score = 0.
start_time = time.clock()
if mode == 'train':
done_looping = False
else:
done_looping = True
epoch = 0
while (epoch < training_epochs) and (not done_looping):
epoch = epoch + 1
for minibatch_index in xrange(n_train_batches):
minibatch_avg_cost = train_fn(minibatch_index)
## save the parameters
if mode == 'train':
get_params = theano.function(inputs=[],
outputs=[sda.sigmoid_layers[0].W, sda.sigmoid_layers[0].b,
sda.sigmoid_layers[1].W, sda.sigmoid_layers[1].b,
sda.sigmoid_layers[2].W, sda.sigmoid_layers[2].b,
sda.logLayer.W, sda.logLayer.b])
save_parameters(get_params(), 'SdA')
iter = (epoch - 1) * n_train_batches + minibatch_index
if (iter + 1) % validation_frequency == 0:
validation_losses = validate_model()
this_validation_loss = numpy.mean(validation_losses)
print('epoch %i, minibatch %i/%i, validation error %f %%' %
(epoch, minibatch_index + 1, n_train_batches,
this_validation_loss * 100.))
# if we got the best validation score until now
if this_validation_loss < best_validation_loss:
#improve patience if loss improvement is good enough
if (this_validation_loss < best_validation_loss *
improvement_threshold):
patience = max(patience, iter * patience_increase)
# save best validation score and iteration number
best_validation_loss = this_validation_loss
best_iter = iter
if patience <= iter:
done_looping = True
break
if mode == 'test':
print 'predicting the labels...'
pred_labels = [[0 for j in xrange(batch_size)] for i in xrange(n_test_batches)]
for i in xrange(n_test_batches):
print str(i+1), '/', str(n_test_batches)
pred_labels[i] = get_test_labels(i)
writer = csv.writer(file('result/SdA.csv', 'w'))
row = 1
print 'output test labels...'
for i in xrange(len(pred_labels)):
print str(i+1), '/', str(len(pred_labels))
for j in xrange(len(pred_labels[i])):
writer.writerow([row, pred_labels[i][j]])
row += 1
end_time = time.clock()
print(('Optimization complete with best validation score of %f %%,'
'with test performance %f %%') %
(best_validation_loss * 100., test_score * 100.))
print >> sys.stderr, ('The training code for file ' +
os.path.split(__file__)[1] +
' ran for %.2fm' % ((end_time - start_time) / 60.))
def evaluate_lenet5(learning_rate=0.1, learning_rate2=0.05, learning_rate3=0.01, n_epochs=200,
dataset='cifar-10-batches-py',
nkerns=[6, 16], batch_size=20, mode='train', amount='full'): # nkerns coule be ok with [10, 50]
""" Demonstrates lenet on MNIST dataset
:type learning_rate: float
:param learning_rate: learning rate used (factor for the stochastic
gradient)
:type n_epochs: int
:param n_epochs: maximal number of epochs to run the optimizer
:type dataset: string
:param dataset: path to the dataset used for training /testing (MNIST here)
:type nkerns: list of ints
:param nkerns: number of kernels on each layer
"""
#learning_rate = theano.shared(value=learning_rate, borrow=True)
rng = numpy.random.RandomState(23455)
datasets = load_data(dataset, mode=mode, amount=amount)
if mode == 'train':
train_set_x, train_set_y = datasets[0]
valid_set_x, valid_set_y = datasets[1]
else:
test_set_x, test_set_y = datasets[0]
# compute number of minibatches for training, validation and testing
if mode == 'train':
n_train_batches = train_set_x.get_value(borrow=True).shape[0]
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_train_batches /= batch_size
n_valid_batches /= batch_size
else:
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
# allocate symbolic variables for the data
index = T.lscalar() # index to a [mini]batch
x = T.matrix('x') # the data is presented as rasterized images
y = T.ivector('y') # the labels are presented as 1D vector of
# [int] labels
ishape = (32, 32) # this is the size of CIFIA-10 images (gray-scaled)
######################
# BUILD ACTUAL MODEL #
######################
print '... building the model'
# Reshape matrix of rasterized images of shape (batch_size,32*32)
# to a 4D tensor, compatible with our LeNetConvPoolLayer
layer0_input = x.reshape((batch_size, 1, 32, 32))
# Construct the first convolutional pooling layer:
# filtering reduces the image size to (32-5+1,32-5+1)=(28,28)
# maxpooling reduces this further to (28/2,28/2) = (14,14)
# 4D output tensor is thus of shape (batch_size,nkerns[0],14,14)
layer0 = LeNetConvPoolLayer(rng, input=layer0_input,
image_shape=(batch_size, 1, 32, 32),
filter_shape=(nkerns[0], 1, 5, 5), poolsize=(2, 2))
# Construct the second convolutional pooling layer
# filtering reduces the image size to (14-5+1,14-5+1)=(10,10)
# maxpooling reduces this further to (10/2,10/2) = (5,5)
# 4D output tensor is thus of shape (nkerns[0],nkerns[1],5,5)
layer1 = LeNetConvPoolLayer(rng, input=layer0.output,
image_shape=(batch_size, nkerns[0], 14, 14),
filter_shape=(nkerns[1], nkerns[0], 5, 5), poolsize=(2, 2))
# the HiddenLayer being fully-connected, it operates on 2D matrices of
# shape (batch_size,num_pixels) (i.e matrix of rasterized images).
# This will generate a matrix of shape (20,50*5*5) = (20,1250) <-??
layer2_input = layer1.output.flatten(2)
# construct a fully-connected sigmoidal layer
layer2 = HiddenLayer(rng, input=layer2_input, n_in=nkerns[1] * 5 * 5,
n_out=500, activation=T.tanh)
# classify the values of the fully-connected sigmoidal layer
layer3 = LogisticRegression(input=layer2.output, n_in=500, n_out=10)
## load the saved parameters
if mode == 'test':
learned_params = unpickle('params/convolutional_mlp_gray.pkl')
# the cost we minimize during training is the NLL of the model
cost = layer3.negative_log_likelihood(y)
# create a function to compute the mistakes that are made by the model
if mode == 'test':
test_model = theano.function([index], layer3.errors(y),
givens={
x: test_set_x[index * batch_size: (index + 1) * batch_size],
y: test_set_y[index * batch_size: (index + 1) * batch_size]})
else:
validate_model = theano.function([index], layer3.errors(y),
givens={
x: valid_set_x[index * batch_size: (index + 1) * batch_size],
y: valid_set_y[index * batch_size: (index + 1) * batch_size]})
check_label = theano.function(inputs=[index],
outputs=layer3.y_pair(y),
givens={
x: train_set_x[index * batch_size: (index + 1) * batch_size],
y: train_set_y[index * batch_size: (index + 1) * batch_size]})
# create a function to get the labels predicted by the model
if mode == 'test':
get_test_labels = theano.function([index], layer3.y_pred,
givens={
x: test_set_x[index * batch_size: (index + 1) * batch_size],
layer0.W: learned_params[0],
layer0.b: learned_params[1],
layer1.W: learned_params[2],
layer1.b: learned_params[3],
layer2.W: learned_params[4],
layer2.b: learned_params[5],
layer3.W: learned_params[6],
layer3.b: learned_params[7]})
if mode == 'train':
# create a list of all model parameters to be fit by gradient descent
params = layer3.params + layer2.params + layer1.params + layer0.params
# create a list of gradients for all model parameters
grads = T.grad(cost, params)
# train_model is a function that updates the model parameters by
# SGD Since this model has many parameters, it would be tedious to
# manually create an update rule for each model parameter. We thus
# create the updates list by automatically looping over all
# (params[i],grads[i]) pairs.
if mode == 'train':
updates = []
for param_i, grad_i in zip(params, grads):
updates.append((param_i, param_i - learning_rate * grad_i))
updates2 = []
for param_i, grad_i in zip(params, grads):
updates2.append((param_i, param_i - learning_rate2 * grad_i))
updates3 = []
for param_i, grad_i in zip(params, grads):
updates3.append((param_i, param_i - learning_rate3 * grad_i))
if mode == 'train':
train_model = theano.function([index], cost, updates=updates,
givens={
x: train_set_x[index * batch_size: (index + 1) * batch_size],
y: train_set_y[index * batch_size: (index + 1) * batch_size]})
train_model2 = theano.function([index], cost, updates=updates2,
givens={
x: train_set_x[index * batch_size: (index + 1) * batch_size],
y: train_set_y[index * batch_size: (index + 1) * batch_size]})
train_model3 = theano.function([index], cost, updates=updates3,
givens={
x: train_set_x[index * batch_size: (index + 1) * batch_size],
y: train_set_y[index * batch_size: (index + 1) * batch_size]})
###############
# TRAIN MODEL #
###############
print '... training the model'
# early-stopping parameters
if mode == 'train':
patience = 10000 # look as this many examples regardless
patience_increase = 2 # wait this much longer when a new best is
# found
improvement_threshold = 0.999 # a relative improvement of this much is
# considered significant
validation_frequency = min(n_train_batches, patience / 2)
# go through this many
# minibatche before checking the network
# on the validation set; in this case we
# check every epoch
start_time = time.clock()
if mode == 'train':
best_params = None
best_validation_loss = numpy.inf
best_iter = 0
test_score = 0.
done_looping = False
else:
done_looping = True
epoch = 0
while (epoch < n_epochs) and (not done_looping):
epoch = epoch + 1
for minibatch_index in xrange(n_train_batches):
iter = (epoch - 1) * n_train_batches + minibatch_index
if iter % 100 == 0:
print 'training @ iter = ', iter
if epoch == 1:
cost_ij = train_model(minibatch_index)
elif this_validation_loss < 0.45 and this_validation_loss > 0.35:
cost_ij = train_model2(minibatch_index)
elif this_validation_loss < 0.35:
cost_ij = train_model3(minibatch_index)
else:
cost_ij = train_model(minibatch_index)
## check the contents of predictions occasionaly
'''
if iter % 100 == 0:
[prediction, true_label] = check_label(minibatch_index)
print 'prediction:'
print prediction
print 'true_label:'
print true_label
'''
## save the parameters
if mode == 'train':
get_params = theano.function(inputs=[], outputs=[layer0.W, layer0.b, layer1.W, layer1.b, layer2.W, layer2.b, layer3.W, layer3.b])
save_parameters(get_params(), 'convolutional_mlp_gray')
if (iter + 1) % validation_frequency == 0:
# compute zero-one loss on validation set
validation_losses = [validate_model(i) for i
in xrange(n_valid_batches)]
this_validation_loss = numpy.mean(validation_losses)
print('epoch %i, minibatch %i/%i, validation error %f %%' % \
(epoch, minibatch_index + 1, n_train_batches, \
this_validation_loss * 100.))
# if we got the best validation score until now
if this_validation_loss < best_validation_loss:
#improve patience if loss improvement is good enough
if this_validation_loss < best_validation_loss * \
improvement_threshold:
patience = max(patience, iter * patience_increase)
# save best validation score and iteration number
best_validation_loss = this_validation_loss
best_iter = iter
'''
# test it on the test set
test_losses = [test_model(i) for i in xrange(n_test_batches)]
test_score = numpy.mean(test_losses)
print((' epoch %i, minibatch %i/%i, test error of best '
'model %f %%') %
(epoch, minibatch_index + 1, n_train_batches,
test_score * 100.))
'''
'''
if patience <= iter:
done_looping = True
break
'''
if mode == 'test':
print 'predicting the labels...'
pred_labels = [[0 for j in xrange(batch_size)] for i in xrange(n_test_batches)]
for i in xrange(n_test_batches):
print str(i+1), '/', str(n_test_batches)
pred_labels[i] = get_test_labels(i)
writer = csv.writer(file('result/convolutional_mlp_gray.csv', 'w'))
row = 1
print 'output test labels...'
for i in xrange(len(pred_labels)): # TBF: hard code
print str(i+1), '/', str(len(pred_labels))
for j in xrange(len(pred_labels[i])):
writer.writerow([row, pred_labels[i][j]])
row += 1
end_time = time.clock()
if mode == 'train':
print('Optimization complete.')
print('Best validation score of %f %% obtained at iteration %i,'\
'with test performance %f %%' %
(best_validation_loss * 100., best_iter + 1, test_score * 100.))
print >> sys.stderr, ('The code for file ' +
os.path.split(__file__)[1] +
' ran for %.2fm' % ((end_time - start_time) / 60.))
def test_mlp(learning_rate=0.01, L1_reg=0.00, L2_reg=0.0001, n_epochs=1000,
dataset='cifar-10-batches-py', batch_size=20, n_hidden=500):
"""
Demonstrate stochastic gradient descent optimization for a multilayer
perceptron
This is demonstrated on MNIST.
:type learning_rate: float
:param learning_rate: learning rate used (factor for the stochastic
gradient
:type L1_reg: float
:param L1_reg: L1-norm's weight when added to the cost (see
regularization)
:type L2_reg: float
:param L2_reg: L2-norm's weight when added to the cost (see
regularization)
:type n_epochs: int
:param n_epochs: maximal number of epochs to run the optimizer
:type dataset: string
:param dataset: the path of the MNIST dataset file from
http://www.iro.umontreal.ca/~lisa/deep/data/mnist/mnist.pkl.gz
"""
datasets = load_data(dataset)
train_set_x, train_set_y = datasets[0]
valid_set_x, valid_set_y = datasets[1]
test_set_x, test_set_y = datasets[2]
# compute number of minibatches for training, validation and testing
n_train_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0] / batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0] / batch_size
######################
# BUILD ACTUAL MODEL #
######################
print '... building the model'
# allocate symbolic variables for the data
index = T.lscalar() # index to a [mini]batch
x = T.matrix('x') # the data is presented as rasterized images
y = T.ivector('y') # the labels are presented as 1D vector of
# [int] labels
rng = numpy.random.RandomState(1234)
# construct the MLP class
classifier = MLP(rng=rng, input=x, n_in=32 * 32,
n_hidden=n_hidden, n_out=10)
# the cost we minimize during training is the negative log likelihood of
# the model plus the regularization terms (L1 and L2); cost is expressed
# here symbolically
cost = classifier.negative_log_likelihood(y) \
+ L1_reg * classifier.L1 \
+ L2_reg * classifier.L2_sqr
# compiling a Theano function that computes the mistakes that are made
# by the model on a minibatch
test_model = theano.function(inputs=[index],
outputs=classifier.errors(y),
givens={
x: test_set_x[index * batch_size:(index + 1) * batch_size],
y: test_set_y[index * batch_size:(index + 1) * batch_size]})
validate_model = theano.function(inputs=[index],
outputs=classifier.errors(y),
givens={
x: valid_set_x[index * batch_size:(index + 1) * batch_size],
y: valid_set_y[index * batch_size:(index + 1) * batch_size]})
# compute the gradient of cost with respect to theta (sotred in params)
# the resulting gradients will be stored in a list gparams
gparams = []
for param in classifier.params:
gparam = T.grad(cost, param)
gparams.append(gparam)
# specify how to update the parameters of the model as a list of
# (variable, update expression) pairs
updates = []
# given two list the zip A = [a1, a2, a3, a4] and B = [b1, b2, b3, b4] of
# same length, zip generates a list C of same size, where each element
# is a pair formed from the two lists :
# C = [(a1, b1), (a2, b2), (a3, b3), (a4, b4)]
for param, gparam in zip(classifier.params, gparams):
updates.append((param, param - learning_rate * gparam))
# compiling a Theano function `train_model` that returns the cost, but
# in the same time updates the parameter of the model based on the rules
# defined in `updates`
train_model = theano.function(inputs=[index], outputs=cost,
updates=updates,
givens={
x: train_set_x[index * batch_size:(index + 1) * batch_size],
y: train_set_y[index * batch_size:(index + 1) * batch_size]})
###############
# TRAIN MODEL #
###############
print '... training'
# early-stopping parameters
patience = 10000 # look as this many examples regardless
patience_increase = 2 # wait this much longer when a new best is
# found
improvement_threshold = 0.999 # a relative improvement of this much is
# considered significant
validation_frequency = min(n_train_batches, patience / 2)
# go through this many
# minibatche before checking the network
# on the validation set; in this case we
# check every epoch
best_params = None
best_validation_loss = numpy.inf
best_iter = 0
test_score = 0.
start_time = time.clock()
epoch = 0
done_looping = False
while (epoch < n_epochs) and (not done_looping):
epoch = epoch + 1
for minibatch_index in xrange(n_train_batches):
minibatch_avg_cost = train_model(minibatch_index)
# iteration number
iter = (epoch - 1) * n_train_batches + minibatch_index
if (iter + 1) % validation_frequency == 0:
# compute zero-one loss on validation set
validation_losses = [validate_model(i) for i
in xrange(n_valid_batches)]
this_validation_loss = numpy.mean(validation_losses)
print('epoch %i, minibatch %i/%i, validation error %f %%' %
(epoch, minibatch_index + 1, n_train_batches,
this_validation_loss * 100.))
# if we got the best validation score until now
if this_validation_loss < best_validation_loss:
#improve patience if loss improvement is good enough
if this_validation_loss < best_validation_loss * \
improvement_threshold:
patience = max(patience, iter * patience_increase)
best_validation_loss = this_validation_loss
best_iter = iter
# test it on the test set
test_losses = [test_model(i) for i
in xrange(n_test_batches)]
test_score = numpy.mean(test_losses)
print((' epoch %i, minibatch %i/%i, test error of '
'best model %f %%') %
(epoch, minibatch_index + 1, n_train_batches,
test_score * 100.))
if patience <= iter:
done_looping = True
break
end_time = time.clock()
print(('Optimization complete. Best validation score of %f %% '
'obtained at iteration %i, with test performance %f %%') %
(best_validation_loss * 100., best_iter + 1, test_score * 100.))
print >> sys.stderr, ('The code for file ' +
os.path.split(__file__)[1] +
' ran for %.2fm' % ((end_time - start_time) / 60.))
