def __init__(self,
learning_rate=0.1,
n_epochs=1,
dataset='mnist.pkl.gz',
nkerns=[20, 50],
batch_size=500,
testing=0):
self.data = load_data(dataset)
def main():
print 'Loading data...'
X, y = load_data('ex2data2.txt')
print 'First 10 examples from dataset:'
print '\t\tX\t\t y'
for i in range(10):
print ' {0}\t{1}'.format(X[i], y[i])
print 'Plotting data...'
plot_data(X, y)
print 'Normalizing features...'
X, mu, sigma = feature_normalize(X)
print 'Adding polynomial features...'
X = map_feature(X[:, 0], X[:, 1])
initial_theta = np.zeros((X.shape[1], 1))
lambda_ = 1
print 'Computing cost...'
cost, grad = cost_function_reg(initial_theta, X, y, lambda_)
print 'Optimizing using gradient descent...'
alpha = 0.05
num_iters = 1000
theta = np.zeros((X.shape[1], 1))
theta, J_history = gradient_descent_multi_reg(
X, y, theta, alpha, num_iters, lambda_)
plt.plot(range(1, len(J_history) + 1), J_history, '-b')
plt.xlabel('Number of iterations')
plt.ylabel('Cost J')
plt.show()
print 'Theta computed from gradient descent:'
print theta
chip_data = np.array([[-0.5, 0.7]])
chip_data = (chip_data - mu) / sigma
chip_data = np.insert(chip_data, 0, 1, axis=1)
chip_data = map_feature(chip_data[:, 0], chip_data[:, 1])
prob = sigmoid(chip_data.dot(theta))
print 'For a microchip with test of -0.5 and 0.7, we predict an acceptance of {0}%'.format(prob[0, 0] * 100)
print 'Computing accuracy:\n'
p = predict(theta, X)
accuracy = np.mean(y == p)
print 'Train Accuracy: {0}%'.format(accuracy * 100)
def evaluate_lenet5(learning_rate=0.1, n_epochs=200,
dataset='mnist.pkl.gz',
nkerns=[20, 50], batch_size=500):
""" 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
"""
rng = numpy.random.RandomState(23455)
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]
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_train_batches /= batch_size
n_valid_batches /= batch_size
n_test_batches /= batch_size
# allocate symbolic variables for the data
index = T.lscalar() # index to a [mini]batch
# start-snippet-1
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
######################
# BUILD ACTUAL MODEL #
######################
print '... building the model'
# Reshape matrix of rasterized images of shape (batch_size, 28 * 28)
# to a 4D tensor, compatible with our LeNetConvPoolLayer
# (28, 28) is the size of MNIST images.
layer0_input = x.reshape((batch_size, 1, 28, 28))
# Construct the first convolutional pooling layer:
# filtering reduces the image size to (28-5+1 , 28-5+1) = (24, 24)
# maxpooling reduces this further to (24/2, 24/2) = (12, 12)
# 4D output tensor is thus of shape (batch_size, nkerns[0], 12, 12)
layer0 = LeNetConvPoolLayer(
rng,
input=layer0_input,
image_shape=(batch_size, 1, 28, 28),
filter_shape=(nkerns[0], 1, 5, 5),
poolsize=(2, 2)
)
# Construct the second convolutional pooling layer
# filtering reduces the image size to (12-5+1, 12-5+1) = (8, 8)
# maxpooling reduces this further to (8/2, 8/2) = (4, 4)
# 4D output tensor is thus of shape (batch_size, nkerns[1], 4, 4)
layer1 = LeNetConvPoolLayer(
rng,
input=layer0.output,
image_shape=(batch_size, nkerns[0], 12, 12),
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 (batch_size, nkerns[1] * 4 * 4),
# or (500, 50 * 4 * 4) = (500, 800) with the default values.
layer2_input = layer1.output.flatten(2)
# construct a fully-connected sigmoidal layer
layer2 = HiddenLayer(
rng,
input=layer2_input,
n_in=nkerns[1] * 4 * 4,
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)
# 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
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]
}
)
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]
}
)
# 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.
updates = [
(param_i, param_i - learning_rate * grad_i)
for param_i, grad_i in zip(params, grads)
]
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]
}
)
# end-snippet-1
###############
# 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 = 0.995 # 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
best_validation_loss = numpy.inf
best_iter = 0
test_score = 0.
start_time = timeit.default_timer()
epoch = 0
done_looping = False
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
cost_ij = train_model(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)
# 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
end_time = timeit.default_timer()
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='mnist.pkl.gz', batch_size=20, n_hidden=500, n_hidden_2=50):
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]
# 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
#import ipdb; ipdb.set_trace()
# Build the model
print '... building the model'
# symbolic variables for the data
index = T.lscalar()
x = T.matrix('x')
y = T.ivector('y')
rng = np.random.RandomState(1234)
# construct the MLP class
classifier = MLP(
rng=rng,
input=x,
n_in=28 * 28,
n_hidden=n_hidden,
n_hidden_2=n_hidden_2,
n_out=10
)
# minimize negative log likelihood & regularization terms during training
cost = (
classifier.negative_log_likelihood(y) +
L1_reg * classifier.L1 +
L2_reg * classifier.L2_sqr
)
print 'cost={}'.format(cost)
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 wrt theata
print classifier.params
gparams = [T.grad(cost, param) for param in classifier.params]
print gparams
# Specify how to update the parameters of the model
updates = [
(param, param - learning_rate * gparam) for param, gparam in zip(classifier.params, gparams)
]
# Compile training function
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 the model
print '... training'
patience = 10000
patience_increase = 2
improvement_threshold = 0.995
validation_frequency = min(n_train_batches, patience/2)
best_validation_loss = np.inf
best_iter = 0
test_score = 0.
start_time = timeit.default_timer()
epoch = 0
done_looping = False
print 'Number of minibatches: {}'.format(n_train_batches)
while (epoch < n_epochs) and (not done_looping):
epoch += 1
epoch_start_time = timeit.default_timer()
for minibatch_index in xrange(n_train_batches):
minibatch_avg_cost = train_model(minibatch_index)
iter = (epoch - 1) * n_train_batches + minibatch_index
if (iter + 1) % validation_frequency == 0:
validation_losses = [validate_model(i) for i in xrange(n_valid_batches)]
this_validation_loss = np.mean(validation_losses)
print 'epoch {}, minibatch {}/{}, validation error {} %'.format(
epoch,
minibatch_index + 1,
n_train_batches,
this_validation_loss * 100.
)
# If this is the best validation score up until now
if this_validation_loss < best_validation_loss:
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 against tes set
test_losses = [test_model(i) for i in xrange(n_test_batches)]
test_score = np.mean(test_losses)
print ' epoch {}, minibatch {}/{}, test error of best model {} %'.format(
epoch,
minibatch_index + 1,
n_train_batches,
test_score * 100.
)
if patience <= iter:
done_looping = True
break
epoch_end_time = timeit.default_timer()
print ' epoch {}, ran for {}s'.format(
epoch,
(epoch_end_time - epoch_start_time)
)
end_time = timeit.default_timer()
print 'Optimization complete. Best validation score of {} %'.format(
best_validation_loss * 100.
)
print 'obtained at iteration {}, with test performance {} %'.format(
best_iter + 1,
test_score * 100.
)
def evaluate_lenet5(learning_rate=0.1,
n_epochs=200,
dataset='mnist.pkl.gz',
nkerns=[16, 16, 16],
batch_size=500):
rng = numpy.random.RandomState(32324)
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]
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
index = T.lscalar() # index for each mini batch
train_epoch = T.lscalar()
x = T.matrix('x')
y = T.ivector('y')
# ------------------------------- Building Model ----------------------------------
print "...Building the model"
# output image size = (28-5+1+4)/2 = 14
layer_0_input = x.reshape((batch_size, 1, 28, 28))
layer_0 = LeNetConvPoolLayer(rng,
input=layer_0_input,
image_shape=(batch_size, 1, 28, 28),
filter_shape=(nkerns[0], 1, 5, 5),
poolsize=(2, 2),
border_mode=2)
#output image size = (14-3+1)/2 = 6
layer_1 = LeNetConvPoolLayer(rng,
input=layer_0.output,
image_shape=(batch_size, nkerns[0], 14, 14),
filter_shape=(nkerns[1], nkerns[0], 3, 3),
poolsize=(2, 2))
#output image size = (6-3+1)/2 = 2
layer_2 = LeNetConvPoolLayer(rng,
input=layer_1.output,
image_shape=(batch_size, nkerns[1], 6, 6),
filter_shape=(nkerns[2], nkerns[1], 3, 3),
poolsize=(2, 2))
# make the input to hidden layer 2 dimensional
layer_3_input = layer_2.output.flatten(2)
layer_3 = HiddenLayer(rng,
input=layer_3_input,
n_in=nkerns[2] * 2 * 2,
n_out=200,
activation=T.tanh)
layer_4 = LogReg(input=layer_3.output, n_in=200, n_out=10)
teacher_p_y_given_x = theano.shared(numpy.asarray(
pickle.load(open('prob_best_model.pkl', 'rb')),
dtype=theano.config.floatX),
borrow=True)
#cost = layer_4.neg_log_likelihood(y) + T.mean((teacher_W - layer_4.W)**2)/(2.*(1+epoch*2)) + T.mean((teacher_b-layer_4.b)**2)/(2.*(1+epoch*2))
# import pdb
# pdb.set_trace()
p_y_given_x = T.matrix('p_y_given_x')
e = theano.shared(value=0, name='e', borrow=True)
#cost = layer_4.neg_log_likelihood(y) + 1.0/(e)*T.mean((layer_4.p_y_given_x - p_y_given_x)**2)
cost = layer_4.neg_log_likelihood(
y) + 2.0 / (e) * T.mean(-T.log(layer_4.p_y_given_x) * p_y_given_x -
layer_4.p_y_given_x * T.log(p_y_given_x))
test_model = theano.function(
[index],
layer_4.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(
[index],
layer_4.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]
})
# list of parameters
params = layer_4.params + layer_3.params + layer_2.params + layer_1.params + layer_0.params
grads = T.grad(cost, params)
updates = [(param_i, param_i - learning_rate * grad_i)
for param_i, grad_i in zip(params, grads)]
train_model = theano.function(
[index, train_epoch],
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],
p_y_given_x: teacher_p_y_given_x[index],
e: train_epoch
})
# -----------------------------------------Starting Training ------------------------------
print('..... Training ')
# for early stopping
patience = 10000
patience_increase = 2
improvement_threshold = 0.95
validation_frequency = min(n_train_batches, patience // 2)
best_validation_loss = numpy.inf # initialising loss to be inifinite
best_itr = 0
test_score = 0
start_time = timeit.default_timer()
#epo = theano.shared('epo')
epoch = 0
done_looping = False
while (epoch < n_epochs) and (not done_looping):
epoch = epoch + 1
for minibatch_index in range(n_train_batches):
iter = (epoch - 1) * n_train_batches + minibatch_index
if iter % 100 == 0:
print('training @ iter = ', iter)
cost_ij = train_model(minibatch_index, epoch)
if (iter + 1) % validation_frequency == 0:
# compute loss on validation set
validation_losses = [
validate_model(i) for i in range(n_valid_batches)
]
this_validation_loss = numpy.mean(validation_losses)
# import pdb
# pdb.set_trace()
print('epoch %i, minibatch %i/%i, validation error %f %%' %
(epoch, minibatch_index + 1, n_train_batches,
this_validation_loss * 100.))
# check with best validation score till now
if this_validation_loss < best_validation_loss:
# improve
if this_validation_loss < best_validation_loss * improvement_threshold:
patience = max(patience, iter * patience_increase)
best_validation_loss = this_validation_loss
best_itr = iter
test_losses = [
test_model(i) for i in range(n_test_batches)
]
test_score = numpy.mean(test_losses)
print('epoch %i, minibatch %i/%i, testing error %f %%' %
(epoch, minibatch_index + 1, n_train_batches,
test_score * 100.))
with open('best_model_3layer.pkl', 'wb') as f:
pickle.dump(params, f)
with open('Results_student_3.txt', 'wb') as f2:
f2.write(str(test_score * 100) + '\n')
#if patience <= iter:
# done_looping = True
# break
end_time = timeit.default_timer()
print('Optimization complete')
print(
'Best validation score of %f %% obtained at iteration %i,'
'with test performance %f %%' %
(best_validation_loss * 100., best_itr, test_score * 100))
print('The code ran for %.2fm' % ((end_time - start_time) / 60.))
def test_rbm(learning_rate=0.1, training_epochs=15,
dataset='mnist.pkl.gz', batch_size=20,
n_chains=20, n_samples=10, output_folder='rbm_plots',
n_hidden=500):
"""
Demonstrate how to train and afterwards sample from it using Theano.
This is demonstrated on MNIST.
:param learning_rate: learning rate used for training the RBM
:param training_epochs: number of epochs used for training
:param dataset: path the the pickled dataset
:param batch_size: size of a batch used to train the RBM
:param n_chains: number of parallel Gibbs chains to be used for sampling
:param n_samples: number of samples to plot for each chain
"""
datasets = load_data(dataset)
train_set_x, train_set_y = datasets[0]
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
# 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
rng = numpy.random.RandomState(123)
theano_rng = RandomStreams(rng.randint(2 ** 30))
# initialize storage for the persistent chain (state = hidden
# layer of chain)
persistent_chain = theano.shared(numpy.zeros((batch_size, n_hidden),
dtype=theano.config.floatX),
borrow=True)
# construct the RBM class
rbm = RBM(input=x, n_visible=28 * 28,
n_hidden=n_hidden, numpy_rng=rng, theano_rng=theano_rng)
# get the cost and the gradient corresponding to one step of CD-15
cost, updates = rbm.get_cost_updates(lr=learning_rate,
persistent=persistent_chain, k=15)
#################################
# Training the RBM #
#################################
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
os.chdir(output_folder)
# it is ok for a theano function to have no output
# the purpose of train_rbm is solely to update the RBM parameters
train_rbm = theano.function([index], cost,
updates=updates,
givens={x: train_set_x[index * batch_size:
(index + 1) * batch_size]},
name='train_rbm')
plotting_time = 0.
start_time = time.clock()
# go through training epochs
for epoch in xrange(training_epochs):
# go through the training set
mean_cost = []
for batch_index in xrange(n_train_batches):
mean_cost += [train_rbm(batch_index)]
print 'Training epoch %d, cost is ' % epoch, numpy.mean(mean_cost)
# Plot filters after each training epoch
plotting_start = time.clock()
# Construct image from the weight matrix
image = PIL.Image.fromarray(tile_raster_images(
X=rbm.W.get_value(borrow=True).T,
img_shape=(28, 28), tile_shape=(10, 10),
tile_spacing=(1, 1)))
image.save('filters_at_epoch_%i.png' % epoch)
plotting_stop = time.clock()
plotting_time += (plotting_stop - plotting_start)
end_time = time.clock()
pretraining_time = (end_time - start_time) - plotting_time
print ('Training took %f minutes' % (pretraining_time / 60.))
#################################
# Sampling from the RBM #
#################################
# find out the number of test samples
number_of_test_samples = test_set_x.get_value(borrow=True).shape[0]
# pick random test examples, with which to initialize the persistent chain
test_idx = rng.randint(number_of_test_samples - n_chains)
persistent_vis_chain = theano.shared(numpy.asarray(
test_set_x.get_value(borrow=True)[test_idx:test_idx + n_chains],
dtype=theano.config.floatX))
plot_every = 1000
# define one step of Gibbs sampling (mf = mean-field) define a
# function that does `plot_every` steps before returning the
# sample for plotting
[presig_hids, hid_mfs, hid_samples, presig_vis,
vis_mfs, vis_samples], updates = \
theano.scan(rbm.gibbs_vhv,
outputs_info=[None, None, None, None,
None, persistent_vis_chain],
n_steps=plot_every)
# add to updates the shared variable that takes care of our persistent
# chain :.
updates.update({persistent_vis_chain: vis_samples[-1]})
# construct the function that implements our persistent chain.
# we generate the "mean field" activations for plotting and the actual
# samples for reinitializing the state of our persistent chain
sample_fn = theano.function([], [vis_mfs[-1], vis_samples[-1]],
updates=updates,
name='sample_fn')
# create a space to store the image for plotting ( we need to leave
# room for the tile_spacing as well)
image_data = numpy.zeros((29 * n_samples + 1, 29 * n_chains - 1),
dtype='uint8')
for idx in xrange(n_samples):
# generate `plot_every` intermediate samples that we discard,
# because successive samples in the chain are too correlated
vis_mf, vis_sample = sample_fn()
print ' ... plotting sample ', idx
image_data[29 * idx:29 * idx + 28, :] = tile_raster_images(
X=vis_mf,
img_shape=(28, 28),
tile_shape=(1, n_chains),
tile_spacing=(1, 1))
# construct image
print numpy.shape(image_data)
image = PIL.Image.fromarray(image_data)
image.save('samples.png')
os.chdir('../')
def evaluate_lenet5(learning_rate = 0.10, n_epochs=200, dataset='mnist.pkl.gz',nkerns = [16,16,16,12,12,12], batch_size = 500):
rng = numpy.random.RandomState(32324)
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]
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
index = T.lscalar() # index for each mini batch
train_epoch = T.lscalar('train_epoch')
x = T.matrix('x')
y = T.ivector('y')
# ------------------------------- Building Model ----------------------------------
print "...Building the model"
layer_0_input = x.reshape((batch_size,1,28,28))
# output image size = (28-5+1+)/1 = 24
layer_0 = LeNetConvPoolLayer(rng,input = layer_0_input, image_shape=(batch_size,1,28,28),
filter_shape=(nkerns[0],1,5,5),poolsize=(1,1))
#output image size = (24-3+1) = 22
layer_1 = LeNetConvPoolLayer(rng, input = layer_0.output, image_shape = (batch_size, nkerns[0],24,24),
filter_shape = (nkerns[1],nkerns[0],3,3), poolsize=(1,1) )
#output image size = (22-3+1)/2 = 10
layer_2 = LeNetConvPoolLayer(rng, input = layer_1.output, image_shape = (batch_size, nkerns[1],22,22),
filter_shape = (nkerns[2],nkerns[1],3,3), poolsize=(2,2) )
#output image size = (10-3+1)/2 = 4
layer_3 = LeNetConvPoolLayer(rng, input = layer_2.output, image_shape = (batch_size, nkerns[2],10,10),
filter_shape = (nkerns[3], nkerns[2],3,3), poolsize=(2,2) )
#output image size = (4-3+2+1) = 4
layer_4 = LeNetConvPoolLayer(rng, input = layer_3.output, image_shape = (batch_size, nkerns[3],4,4),
filter_shape = (nkerns[4], nkerns[3],3,3), poolsize=(1,1), border_mode = 1 )
#output image size = (4-3+1)/2 = 2
layer_5 = LeNetConvPoolLayer(rng, input = layer_4.output, image_shape = (batch_size, nkerns[4],4,4),
filter_shape = (nkerns[5], nkerns[4],3,3), poolsize=(2,2), border_mode = 1 )
# make the input to hidden layer 2 dimensional
layer_6_input = layer_5.output.flatten(2)
layer_6 = HiddenLayer(rng,input = layer_6_input, n_in = nkerns[5]*2*2, n_out = 200, activation = T.tanh)
layer_7 = LogReg(input = layer_6.output, n_in=200, n_out = 10)
teacher_p_y_given_x = theano.shared(numpy.asarray(pickle.load(open('prob_best_model.pkl','rb')),dtype =theano.config.floatX), borrow=True)
p_y_given_x = T.matrix('p_y_given_x')
e = theano.shared(value = 0, name = 'e', borrow = True)
cost = layer_7.neg_log_likelihood(y) + 2.0/(e)*T.mean(-T.log(layer_7.p_y_given_x)*p_y_given_x - layer_7.p_y_given_x*T.log(p_y_given_x))
tg = theano.shared(numpy.asarray(pickle.load(open('modified_guided_data.pkl','rb')),dtype =theano.config.floatX), borrow=True)
guiding_weights = T.tensor4('guiding_weights')
#guide_cost = T.mean(-T.log(layer_3.output)*guiding_weights - layer_3.output*T.log(guiding_weights))
guide_cost = T.mean((layer_3.output-guiding_weights)**2)
test_model = theano.function([index],layer_7.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([index],layer_7.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]
})
# list of parameters
params = layer_7.params + layer_6.params + layer_5.params + layer_4.params + layer_3.params + layer_2.params + layer_1.params + layer_0.params
params_gl = layer_3.params + layer_2.params + layer_1.params + layer_0.params
# import pdb
# pdb.set_trace()
grads_gl = T.grad(guide_cost,params_gl)
updates_gl = [ (param_i,param_i-learning_rate/10*grad_i) for param_i,grad_i in zip(params_gl,grads_gl) ]
grads = T.grad(cost,params)
updates = [ (param_i, param_i-learning_rate*grad_i) for param_i, grad_i in zip(params,grads) ]
train_model = theano.function([index,train_epoch],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],
p_y_given_x: teacher_p_y_given_x[index],
e: train_epoch
})
train_till_guided_layer = theano.function([index],guide_cost,updates = updates_gl,
givens={
x: train_set_x[index*batch_size:(index+1)*batch_size],
y: train_set_y[index*batch_size:(index+1)*batch_size],
guiding_weights : tg[index]
},on_unused_input='ignore')
# -----------------------------------------Starting Training ------------------------------
print ('..... Training ' )
# for early stopping
patience = 10000
patience_increase = 2
improvement_threshold = 0.95
validation_frequency = min(n_train_batches, patience//2)
best_validation_loss = numpy.inf # initialising loss to be inifinite
best_itr = 0
test_score = 0
start_time = timeit.default_timer()
epoch = 0
done_looping = False
while (epoch < n_epochs) and (not done_looping) :
epoch = epoch+1
for minibatch_index in range(n_train_batches):
iter = (epoch - 1)*n_train_batches + minibatch_index
if iter%100==0:
print ('training @ iter = ', iter)
if epoch < n_epochs/5:
cost_ij_guided = train_till_guided_layer(minibatch_index)
cost_ij = train_model(minibatch_index,epoch)
if(iter +1)%validation_frequency ==0:
# compute loss on validation set
validation_losses = [validate_model(i) for i in range(n_valid_batches)]
this_validation_loss = numpy.mean(validation_losses)
# import pdb
# pdb.set_trace()
with open('Student_6_terminal_out','a+') as f_:
f_.write('epoch %i, minibatch %i/%i, validation error %f %% \n' %(epoch,minibatch_index+1,n_train_batches,this_validation_loss*100. ))
# check with best validation score till now
if this_validation_loss<best_validation_loss:
# improve
if this_validation_loss < best_validation_loss * improvement_threshold:
patience = max(patience, iter*patience_increase)
best_validation_loss = this_validation_loss
best_itr = iter
test_losses = [test_model(i) for i in range(n_test_batches)]
test_score = numpy.mean(test_losses)
with open('Student_6_terminal_out','a+') as f_:
f_.write('epoch %i, minibatch %i/%i, testing error %f %%\n' %(epoch, minibatch_index+1,n_train_batches,test_score*100.))
with open('best_model_7layer.pkl', 'wb') as f:
pickle.dump(params, f)
with open('Results_student_6.txt', 'wb') as f1:
f1.write(str(test_score*100)+'\n')
#if patience <= iter:
# done_looping = True
# break
end_time = timeit.default_timer()
with open('Student_6_terminal_out','a+') as f_:
f_.write('Optimization complete\n')
f_.write('Best validation score of %f %% obtained at iteration %i, with test performance %f %% \n' % (best_validation_loss*100., best_itr, test_score*100 ))
f_.write('The code ran for %.2fm \n' %((end_time - start_time)/60.))
def test_mlp(learning_rate=0.01,
L1_reg=0.00,
L2_reg=0.0001,
n_epochs=1000,
dataset='mnist.pkl.gz',
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=28 * 28,
n_hidden=n_hidden,
n_out=10)
# start-snippet-4
# 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)
# end-snippet-4
# 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]
})
# start-snippet-5
# compute the gradient of cost with respect to theta (sotred in params)
# the resulting gradients will be stored in a list gparams
gparams = [T.grad(cost, param) for param in classifier.params]
# specify how to update the parameters of the model as a list of
# (variable, update expression) pairs
# given two lists of the same length, A = [a1, a2, a3, a4] and
# B = [b1, b2, b3, b4], 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)]
updates = [(param, param - learning_rate * gparam)
for param, gparam in zip(classifier.params, gparams)]
# 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]
})
# end-snippet-5
###############
# 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 = 0.995 # 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
best_validation_loss = numpy.inf
best_iter = 0
test_score = 0.
start_time = timeit.default_timer()
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 = timeit.default_timer()
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.))
def evaluate_lenet5(learning_rate = 0.1, n_epochs=200, dataset='mnist.pkl.gz',nkerns = [20,50], batch_size = 500 , testing =0):
rng = numpy.random.RandomState(32324)
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]
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
index = T.lscalar() # index for each mini batch
x = T.matrix('x')
y = T.ivector('y')
# ------------------------------- Building Model ----------------------------------
if testing ==0:
print "...Building the model"
# output image size = (28-5+1)/2 = 12
layer_0_input = x.reshape((batch_size,1,28,28))
layer_0 = LeNetConvPoolLayer(rng,input = layer_0_input, image_shape=(batch_size,1,28,28),filter_shape=(nkerns[0],1,5,5),poolsize=(2,2))
#output image size = (12-5+1)/2 = 4
layer_1 = LeNetConvPoolLayer(rng, input = layer_0.output, image_shape = (batch_size, nkerns[0],12,12),
filter_shape = (nkerns[1],nkerns[0],5,5), poolsize=(2,2) )
# make the input to hidden layer 2 dimensional
layer_2_input = layer_1.output.flatten(2)
layer_2 = HiddenLayer(rng,input = layer_2_input, n_in = nkerns[1]*4*4, n_out = 500, activation = T.tanh)
layer_3 = LogReg(input = layer_2.output, n_in=500, n_out = 10)
cost = layer_3.neg_log_likelihood(y)
test_model = theano.function([index],layer_3.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([index],layer_3.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]
})
train_predic = theano.function([index], layer_3.prob_y_given_x(),
givens={
x: train_set_x[index*batch_size:(index+1)*batch_size]
})
# list of parameters
layer_guided = theano.function([index], layer_1.output,
givens={
x: train_set_x[index*batch_size:(index+1)*batch_size]
})
params = layer_3.params + layer_2.params + layer_1.params + layer_0.params
grads = T.grad(cost,params)
updates = [ (param_i, param_i-learning_rate*grad_i) for param_i, grad_i in zip(params,grads) ]
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]
})
# -----------------------------------------Starting Training ------------------------------
if testing ==0:
print ('..... Training ' )
# for early stopping
patience = 10000
patience_increase = 2
improvement_threshold = 0.95
validation_frequency = min(n_train_batches, patience//2)
best_validation_loss = numpy.inf # initialising loss to be inifinite
best_itr = 0
test_score = 0
start_time = timeit.default_timer()
epoch = 0
done_looping = False
while (epoch < n_epochs) and (not done_looping) and testing ==0:
epoch = epoch+1
for minibatch_index in range(n_train_batches):
iter = (epoch - 1)*n_train_batches + minibatch_index
if iter%100 ==0:
print ('training @ iter = ', iter)
cost_ij = train_model(minibatch_index)
if(iter +1)%validation_frequency ==0:
# compute loss on validation set
validation_losses = [validate_model(i) for i in range(n_valid_batches)]
this_validation_loss = numpy.mean(validation_losses)
# import pdb
# pdb.set_trace()
print ('epoch %i, minibatch %i/%i, validation error %f %%' %(epoch,minibatch_index+1,n_train_batches,this_validation_loss*100. ))
# check with best validation score till now
if this_validation_loss<best_validation_loss:
# improve
# if this_validation_loss < best_validation_loss * improvement_threshold:
# patience = max(patience, iter*patience_increase)
best_validation_loss = this_validation_loss
best_itr = iter
test_losses = [test_model(i) for i in range(n_test_batches)]
test_score = numpy.mean(test_losses)
print ('epoch %i, minibatch %i/%i, testing error %f %%' %(epoch, minibatch_index+1,n_train_batches,test_score*100.))
with open('best_model.pkl', 'wb') as f:
pickle.dump(params, f)
with open('Results_teacher.txt','wb') as f2:
f2.write(str(test_score*100) + '\n')
p_y_given_x = [train_predic(i) for i in range(n_train_batches)]
with open ('prob_best_model.pkl','wb') as f1:
pickle.dump(p_y_given_x,f1)
# if patience <= iter:
# done_looping = True
# break
layer_2_op_dump = [layer_guided(i) for i in range(n_train_batches)]
with open ('layer_guided.pkl','wb') as lg:
pickle.dump(layer_2_op_dump,lg)
end_time = timeit.default_timer()
# p_y_given_x = [train_model(i) for i in range(n_train_batches)]
# with open ('prob_best_model.pkl') as f:
# pickle.dump(p_y_given_x)
if testing ==0 :
print ('Optimization complete')
print ('Best validation score of %f %% obtained at iteration %i,'
'with test performance %f %%' % (best_validation_loss*100., best_itr, test_score*100 ))
print('The code 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='mnist.pkl.gz', 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=28 * 28,
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 = 0.995 # 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
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.))
