def __init__(self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
n_outs=1,
corruption_levels=[0.1, 0.1]):
""" This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the sdA
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
:type corruption_levels: list of float
:param corruption_levels: amount of corruption to use for each
layer
"""
self.sigmoid_layers = []
self.dA_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2**30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.matrix('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
self.params.extend(sigmoid_layer.params)
# Construct a denoising autoencoder that shared weights with this
# layer
dA_layer = dA(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
bhid=sigmoid_layer.b)
self.dA_layers.append(dA_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs)
self.get_prediction = theano.function(inputs=[self.x],
outputs=[self.logLayer.y_pred])
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
# self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.sigmoid_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question... but we are
# going to only declare that the parameters of the
# sigmoid_layers are parameters of the DBN. The visible
# biases in the RBM are parameters of those RBMs, but not
# of the DBN.
self.params.extend(sigmoid_layer.params)
# Construct an RBM that shared weights with this layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
hbias=sigmoid_layer.b)
self.rbm_layers.append(rbm_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs)
self.get_prediction = theano.function(
inputs = [self.x],
outputs = [self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.logLayer.p_y_given_x]
)
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
class SdA(object):
"""Stacked denoising auto-encoder class (SdA)
A stacked denoising autoencoder model is obtained by stacking several
dAs. The hidden layer of the dA at layer `i` becomes the input of
the dA at layer `i+1`. The first layer dA gets as input the input of
the SdA, and the hidden layer of the last dA represents the output.
Note that after pretraining, the SdA is dealt with as a normal MLP,
the dAs are only used to initialize the weights.
"""
def __init__(self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
n_outs=1,
corruption_levels=[0.1, 0.1]):
""" This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the sdA
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
:type corruption_levels: list of float
:param corruption_levels: amount of corruption to use for each
layer
"""
self.sigmoid_layers = []
self.dA_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2**30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.matrix('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
self.params.extend(sigmoid_layer.params)
# Construct a denoising autoencoder that shared weights with this
# layer
dA_layer = dA(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
bhid=sigmoid_layer.b)
self.dA_layers.append(dA_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs)
self.get_prediction = theano.function(inputs=[self.x],
outputs=[self.logLayer.y_pred])
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
# self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size):
''' Generates a list of functions, each of them implementing one
step in trainnig the dA corresponding to the layer with same index.
The function will require as input the minibatch index, and to train
a dA you just need to iterate, calling the corresponding function on
all minibatch indexes.
:type train_set_x: theano.tensor.TensorType
:param train_set_x: Shared variable that contains all datapoints used
for training the dA
:type batch_size: int
:param batch_size: size of a [mini]batch
:type learning_rate: float
:param learning_rate: learning rate used during training for any of
the dA layers
'''
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
corruption_level = T.scalar('corruption') # % of corruption to use
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
for dA in self.dA_layers:
# get the cost and the updates list
cost, updates = dA.get_cost_updates(corruption_level,
learning_rate)
# compile the theano function
fn = theano.function(
inputs=[
index,
theano.Param(corruption_level, default=0.2),
theano.Param(learning_rate, default=0.1)
],
outputs=cost,
updates=updates,
givens={self.x: train_set_x[batch_begin:batch_end]})
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns
def build_finetune_functions(self, dataset, batch_size, learning_rate):
'''Generates a function `train` that implements one step of
finetuning, a function `validate` that computes the error on
a batch from the validation set, and a function `test` that
computes the error on a batch from the testing set
:type datasets: list of pairs of theano.tensor.TensorType
:param datasets: It is a list that contain all the datasets;
the has to contain three pairs, `train`,
`valid`, `test` in this order, where each pair
is formed of two Theano variables, one for the
datapoints, the other for the labels
:type batch_size: int
:param batch_size: size of a minibatch
:type learning_rate: float
:param learning_rate: learning rate used during finetune stage
'''
train_set_x, train_set_y = theano.shared(
dataset.phase2['train']['x']), theano.shared(
dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(
dataset.phase2['valid']['x']), theano.shared(
dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(
dataset.phase2['test']['x']), theano.shared(
dataset.phase2['test']['y'])
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
# compute the gradients with respect to the model parameters
gparams = T.grad(self.finetune_cost, self.params)
# compute list of fine-tuning updates
updates = []
for param, gparam in zip(self.params, gparams):
updates.append((param, param - gparam * learning_rate))
train_fn = theano.function(
inputs=[index],
outputs=self.finetune_cost,
updates=updates,
givens={
self.x:
train_set_x[index * batch_size:(index + 1) * batch_size],
self.y:
train_set_y[index * batch_size:(index + 1) * batch_size]
},
name='train')
test_score_i = theano.function(
[index],
self.errors,
givens={
self.x:
test_set_x[index * batch_size:(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:(index + 1) * batch_size]
},
name='test')
valid_score_i = theano.function(
[index],
self.errors,
givens={
self.x:
valid_set_x[index * batch_size:(index + 1) * batch_size],
self.y:
valid_set_y[index * batch_size:(index + 1) * batch_size]
},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score
def __init__(
self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
hidden_recurrent=150,
n_outs=1,
y_type=1,
):
self.sigmoid_layers = []
self.rnnrbm_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix("x") # the data is presented as rasterized images
if y_type == 0:
self.y = T.matrix("y") # the labels are presented as 1D vector of
else:
self.y = T.ivector("y") # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
# Construct a denoising autoencoder that shared weights with this
# layer
if i == 0:
rnnrbm_layer = RnnRbm(
n_visible=input_size,
input=layer_input,
n_hidden=hidden_layers_sizes[i],
n_hidden_recurrent=hidden_recurrent,
lr=0.001,
y_type=y_type,
)
self.rnnrbm_layers.append(rnnrbm_layer)
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
W=rnnrbm_layer.W,
b=rnnrbm_layer.bh_t,
)
else:
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
)
rbm_layer = RBM(
numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
hbias=sigmoid_layer.b,
y_type=y_type,
)
self.rbm_layers.append(rbm_layer)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
self.params.extend(sigmoid_layer.params)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output, n_in=hidden_layers_sizes[-1], n_out=n_outs, y_type=y_type
)
self.get_prediction = theano.function(inputs=[self.x], outputs=[self.logLayer.y_pred])
self.get_py = theano.function(inputs=[self.x], outputs=[self.logLayer.p_y_given_x])
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
if y_type == 0:
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
class DBN(object):
"""Deep Belief Network
A deep belief network is obtained by stacking several RBMs on top of each
other. The hidden layer of the RBM at layer `i` becomes the input of the
RBM at layer `i+1`. The first layer RBM gets as input the input of the
network, and the hidden layer of the last RBM represents the output. When
used for classification, the DBN is treated as a MLP, by adding a logistic
regression layer on top.
"""
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.sigmoid_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question... but we are
# going to only declare that the parameters of the
# sigmoid_layers are parameters of the DBN. The visible
# biases in the RBM are parameters of those RBMs, but not
# of the DBN.
self.params.extend(sigmoid_layer.params)
# Construct an RBM that shared weights with this layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
hbias=sigmoid_layer.b)
self.rbm_layers.append(rbm_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs)
self.get_prediction = theano.function(
inputs = [self.x],
outputs = [self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.logLayer.p_y_given_x]
)
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size, k):
'''Generates a list of functions, for performing one step of
gradient descent at a given layer. The function will require
as input the minibatch index, and to train an RBM you just
need to iterate, calling the corresponding function on all
minibatch indexes.
:type train_set_x: theano.tensor.TensorType
:param train_set_x: Shared var. that contains all datapoints used
for training the RBM
:type batch_size: int
:param batch_size: size of a [mini]batch
:param k: number of Gibbs steps to do in CD-k / PCD-k
'''
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
for rbm in self.rbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost, updates = rbm.get_cost_updates(learning_rate,
persistent=None, k=k)
# compile the theano function
fn = theano.function(inputs=[index,
theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x:
train_set_x[batch_begin:batch_end]})
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns
def build_finetune_functions(self, dataset, batch_size, learning_rate):
'''Generates a function `train` that implements one step of
finetuning, a function `validate` that computes the error on a
batch from the validation set, and a function `test` that
computes the error on a batch from the testing set
:type datasets: list of pairs of theano.tensor.TensorType
:param datasets: It is a list that contain all the datasets;
the has to contain three pairs, `train`,
`valid`, `test` in this order, where each pair
is formed of two Theano variables, one for the
datapoints, the other for the labels
:type batch_size: int
:param batch_size: size of a minibatch
:type learning_rate: float
:param learning_rate: learning rate used during finetune stage
'''
train_set_x, train_set_y = theano.shared(dataset.phase2['train']['x']), theano.shared(dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(dataset.phase2['valid']['x']), theano.shared(dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(dataset.phase2['test']['x']), theano.shared(dataset.phase2['test']['y'])
train_set_y = T.cast(train_set_y,'int32')
valid_set_y = T.cast(valid_set_y,'int32')
test_set_y = T.cast(test_set_y,'int32')
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
# compute the gradients with respect to the model parameters
gparams = T.grad(self.finetune_cost, self.params)
# compute list of fine-tuning updates
updates = []
for param, gparam in zip(self.params, gparams):
updates.append((param, param - gparam * learning_rate))
train_fn = theano.function(inputs=[index],
outputs=self.finetune_cost,
updates=updates,
givens={self.x: train_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: train_set_y[index * batch_size:
(index + 1) * batch_size]},
name='train')
test_score_i = theano.function([index], self.errors,
givens={self.x: test_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:
(index + 1) * batch_size]},
name='test')
valid_score_i = theano.function([index], self.errors,
givens={self.x: valid_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: valid_set_y[index * batch_size:
(index + 1) * batch_size]},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500], hidden_recurrent=150, n_outs=1):
self.sigmoid_layers = []
self.rnnrbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.ivector('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
# Construct a denoising autoencoder that shared weights with this
# layer
if i==0:
rnnrbm_layer = RnnRbm(n_visible=input_size,
input=layer_input, n_hidden=hidden_layers_sizes[i],
n_hidden_recurrent=hidden_recurrent,lr=0.001)
self.rnnrbm_layers.append(rnnrbm_layer)
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
W=rnnrbm_layer.W, b=rnnrbm_layer.bh_t)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
else :
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid
)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
self.params.extend(sigmoid_layer.params)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1], n_out=n_outs)
self.get_prediction = theano.function(
inputs=[self.x],
outputs=[self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs=[self.x],
outputs=[self.logLayer.p_y_given_x]
)
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
class RNNRBM_DBN(object):
def __init__(
self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
hidden_recurrent=150,
n_outs=1,
y_type=1,
):
self.sigmoid_layers = []
self.rnnrbm_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix("x") # the data is presented as rasterized images
if y_type == 0:
self.y = T.matrix("y") # the labels are presented as 1D vector of
else:
self.y = T.ivector("y") # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
# Construct a denoising autoencoder that shared weights with this
# layer
if i == 0:
rnnrbm_layer = RnnRbm(
n_visible=input_size,
input=layer_input,
n_hidden=hidden_layers_sizes[i],
n_hidden_recurrent=hidden_recurrent,
lr=0.001,
y_type=y_type,
)
self.rnnrbm_layers.append(rnnrbm_layer)
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
W=rnnrbm_layer.W,
b=rnnrbm_layer.bh_t,
)
else:
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
)
rbm_layer = RBM(
numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
hbias=sigmoid_layer.b,
y_type=y_type,
)
self.rbm_layers.append(rbm_layer)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
self.params.extend(sigmoid_layer.params)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output, n_in=hidden_layers_sizes[-1], n_out=n_outs, y_type=y_type
)
self.get_prediction = theano.function(inputs=[self.x], outputs=[self.logLayer.y_pred])
self.get_py = theano.function(inputs=[self.x], outputs=[self.logLayer.p_y_given_x])
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
if y_type == 0:
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size, k):
# index to a [mini]batch
index = T.lscalar("index") # index to a minibatch
learning_rate = T.scalar("lr") # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
pretrain_f_bh_t = []
for rnnrbm in self.rnnrbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost, updates, bh_t, updates_bh_t = rnnrbm.get_cost_updates(learning_rate)
# compile the theano function
fn = theano.function(
inputs=[index, theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x: train_set_x[batch_begin:batch_end]},
)
# append `fn` to the list of functions
f_bh_t = theano.function(
inputs=[index], outputs=bh_t, updates=updates_bh_t, givens={self.x: train_set_x[batch_begin:batch_end]}
)
pretrain_fns.append(fn)
pretrain_f_bh_t.append(f_bh_t)
for rbm in self.rbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost, updates = rbm.get_cost_updates(learning_rate, persistent=None, k=k)
# compile the theano function
fn = theano.function(
inputs=[index, theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x: train_set_x[batch_begin:batch_end]},
)
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns, pretrain_f_bh_t
def build_finetune_functions(self, dataset, batch_size, learning_rate, y_type):
train_set_x, train_set_y = (
theano.shared(dataset.phase2["train"]["x"]),
theano.shared(dataset.phase2["train"]["y"]),
)
valid_set_x, valid_set_y = (
theano.shared(dataset.phase2["valid"]["x"]),
theano.shared(dataset.phase2["valid"]["y"]),
)
test_set_x, test_set_y = theano.shared(dataset.phase2["test"]["x"]), theano.shared(dataset.phase2["test"]["y"])
if not y_type == 0:
train_set_y = T.cast(train_set_y, "int32")
valid_set_y = T.cast(valid_set_y, "int32")
test_set_y = T.cast(test_set_y, "int32")
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar("index") # index to a [mini]batch
# compute the gradients with respect to the model parameters
gparams = T.grad(self.finetune_cost, self.params)
# compute list of fine-tuning updates
updates = []
for param, gparam in zip(self.params, gparams):
updates.append((param, param - gparam * learning_rate))
train_fn = theano.function(
inputs=[index],
outputs=self.finetune_cost,
updates=updates,
givens={
self.x: train_set_x[index * batch_size : (index + 1) * batch_size],
self.y: train_set_y[index * batch_size : (index + 1) * batch_size],
},
name="train",
)
test_score_i = theano.function(
[index],
self.errors,
givens={
self.x: test_set_x[index * batch_size : (index + 1) * batch_size],
self.y: test_set_y[index * batch_size : (index + 1) * batch_size],
},
name="test",
)
valid_score_i = theano.function(
[index],
self.errors,
givens={
self.x: valid_set_x[index * batch_size : (index + 1) * batch_size],
self.y: valid_set_y[index * batch_size : (index + 1) * batch_size],
},
name="valid",
)
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
train_set_x, train_set_y, valid_set_x, valid_set_y, test_set_x, test_set_y = "", "", "", "", "", ""
return train_fn, valid_score, test_score
def __init__(self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
n_outs=1,
y_type=1,
gbrbm=False,
dropout=False,
activation_function=None):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.dropout = dropout
self.sigmoid_layers = []
self.dropout_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
rectified_linear_activation = lambda x: T.maximum(0.0, x)
# activation_function = T.nnet.sigmoid
if not activation_function:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
else:
if activation_function == 'ReLU':
print 'ReLU'
self.activation_function = rectified_linear_activation
else:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2**30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
if y_type == 0:
self.y = T.matrix('y') # the labels are presented as 1D vector
else:
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
if dropout:
dropout_layer_input = _dropout_from_layer(numpy_rng,
self.x,
p=0.2)
else:
layer_input = self.sigmoid_layers[-1].output
if dropout:
dropout_layer_input = self.dropout_layers[-1].output
if dropout:
print 'Dropout'
dropout_layer = DropoutHiddenLayer(
rng=numpy_rng,
input=dropout_layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
)
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
W=dropout_layer.W * (0.8 if i == 0 else 0.5),
b=dropout_layer.b)
self.dropout_layers.append(dropout_layer)
self.params.extend(dropout_layer.params)
else:
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
self.params.extend(sigmoid_layer.params)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# Construct an RBM that shared weights with this layer
if dropout:
layer = dropout_layer
else:
layer = sigmoid_layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=layer.W,
hbias=layer.b,
y_type=y_type,
gbrbm=gbrbm)
self.rbm_layers.append(rbm_layer)
if dropout:
self.dropout_output_layer = LogisticRegression(
input=self.dropout_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type)
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type,
W=self.dropout_output_layer.W * 0.5,
b=self.dropout_output_layer.b)
else:
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type)
self.get_prediction = theano.function(inputs=[self.x],
outputs=[self.logLayer.y_pred])
self.get_py = theano.function(inputs=[self.x],
outputs=[self.logLayer.p_y_given_x])
if dropout:
self.get_prediction_dropout = theano.function(
inputs=[self.x], outputs=[self.dropout_output_layer.y_pred])
self.get_py = theano.function(
inputs=[self.x],
outputs=[self.dropout_output_layer.p_y_given_x])
if dropout:
self.params.extend(self.dropout_output_layer.params)
else:
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
if y_type == 0:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.squared_error(
self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.negative_log_likelihood(
self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
if dropout:
self.dropout_errors = self.dropout_output_layer.errors(self.y)
self.errors = self.logLayer.errors(self.y)
class RNNRBM_MLP(object):
#class stackRnnRbm(object):
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500], hidden_recurrent=150, n_outs=1):
self.sigmoid_layers = []
self.rnnrbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.ivector('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
# Construct a denoising autoencoder that shared weights with this
# layer
if i==0:
rnnrbm_layer = RnnRbm(n_visible=input_size,
input=layer_input, n_hidden=hidden_layers_sizes[i],
n_hidden_recurrent=hidden_recurrent,lr=0.001)
self.rnnrbm_layers.append(rnnrbm_layer)
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid,
W=rnnrbm_layer.W, b=rnnrbm_layer.bh_t)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
else :
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid
)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
self.params.extend(sigmoid_layer.params)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1], n_out=n_outs)
self.get_prediction = theano.function(
inputs=[self.x],
outputs=[self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs=[self.x],
outputs=[self.logLayer.p_y_given_x]
)
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self,train_set_x,batch_size):
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
pretrain_f_bh_t= []
for rnnrbm in self.rnnrbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost,updates,bh_t,updates_bh_t = rnnrbm.get_cost_updates(learning_rate)
# compile the theano function
fn = theano.function(inputs=[index,
theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x:
train_set_x[batch_begin:batch_end]})
# append `fn` to the list of functions
f_bh_t = theano.function(inputs=[index],
outputs=bh_t,
updates=updates_bh_t,
givens={self.x:
train_set_x[batch_begin:batch_end]})
pretrain_fns.append(fn)
pretrain_f_bh_t.append(f_bh_t)
return pretrain_fns,pretrain_f_bh_t
def build_finetune_functions(self,dataset,batch_size,learning_rate):
train_set_x, train_set_y = theano.shared(dataset.phase2['train']['x']), theano.shared(dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(dataset.phase2['valid']['x']), theano.shared(dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(dataset.phase2['test']['x']), theano.shared(dataset.phase2['test']['y'])
train_set_y = T.cast(train_set_y,'int32')
valid_set_y = T.cast(valid_set_y,'int32')
test_set_y = T.cast(test_set_y,'int32')
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
# compute the gradients with respect to the model parameters
gparams = T.grad(self.finetune_cost, self.params)
# compute list of fine-tuning updates
updates = []
for param, gparam in zip(self.params, gparams):
updates.append((param, param - gparam * learning_rate))
train_fn = theano.function(inputs=[index],
outputs=self.finetune_cost,
updates=updates,
givens={self.x: train_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: train_set_y[index * batch_size:
(index + 1) * batch_size]},
name='train')
test_score_i = theano.function([index], self.errors,
givens={self.x: test_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:
(index + 1) * batch_size]},
name='test')
valid_score_i = theano.function([index], self.errors,
givens={self.x: valid_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: valid_set_y[index * batch_size:
(index + 1) * batch_size]},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score
class DBN(object):
"""Deep Belief Network
A deep belief network is obtained by stacking several RBMs on top of each
other. The hidden layer of the RBM at layer `i` becomes the input of the
RBM at layer `i+1`. The first layer RBM gets as input the input of the
network, and the hidden layer of the last RBM represents the output. When
used for classification, the DBN is treated as a MLP, by adding a logistic
regression layer on top.
"""
def __init__(self,
numpy_rng,
theano_rng=None,
n_ins=784,
hidden_layers_sizes=[500, 500],
n_outs=1,
y_type=1,
gbrbm=False,
dropout=False,
activation_function=None):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.dropout = dropout
self.sigmoid_layers = []
self.dropout_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
rectified_linear_activation = lambda x: T.maximum(0.0, x)
# activation_function = T.nnet.sigmoid
if not activation_function:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
else:
if activation_function == 'ReLU':
print 'ReLU'
self.activation_function = rectified_linear_activation
else:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2**30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
if y_type == 0:
self.y = T.matrix('y') # the labels are presented as 1D vector
else:
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
if dropout:
dropout_layer_input = _dropout_from_layer(numpy_rng,
self.x,
p=0.2)
else:
layer_input = self.sigmoid_layers[-1].output
if dropout:
dropout_layer_input = self.dropout_layers[-1].output
if dropout:
print 'Dropout'
dropout_layer = DropoutHiddenLayer(
rng=numpy_rng,
input=dropout_layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
)
sigmoid_layer = HiddenLayer(
rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
W=dropout_layer.W * (0.8 if i == 0 else 0.5),
b=dropout_layer.b)
self.dropout_layers.append(dropout_layer)
self.params.extend(dropout_layer.params)
else:
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
self.params.extend(sigmoid_layer.params)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# Construct an RBM that shared weights with this layer
if dropout:
layer = dropout_layer
else:
layer = sigmoid_layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=layer.W,
hbias=layer.b,
y_type=y_type,
gbrbm=gbrbm)
self.rbm_layers.append(rbm_layer)
if dropout:
self.dropout_output_layer = LogisticRegression(
input=self.dropout_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type)
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type,
W=self.dropout_output_layer.W * 0.5,
b=self.dropout_output_layer.b)
else:
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs,
y_type=y_type)
self.get_prediction = theano.function(inputs=[self.x],
outputs=[self.logLayer.y_pred])
self.get_py = theano.function(inputs=[self.x],
outputs=[self.logLayer.p_y_given_x])
if dropout:
self.get_prediction_dropout = theano.function(
inputs=[self.x], outputs=[self.dropout_output_layer.y_pred])
self.get_py = theano.function(
inputs=[self.x],
outputs=[self.dropout_output_layer.p_y_given_x])
if dropout:
self.params.extend(self.dropout_output_layer.params)
else:
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
if y_type == 0:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.squared_error(
self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.negative_log_likelihood(
self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
if dropout:
self.dropout_errors = self.dropout_output_layer.errors(self.y)
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size, k):
'''Generates a list of functions, for performing one step of
gradient descent at a given layer. The function will require
as input the minibatch index, and to train an RBM you just
need to iterate, calling the corresponding function on all
minibatch indexes.
:type train_set_x: theano.tensor.TensorType
:param train_set_x: Shared var. that contains all datapoints used
for training the RBM
:type batch_size: int
:param batch_size: size of a [mini]batch
:param k: number of Gibbs steps to do in CD-k / PCD-k
'''
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
for rbm in self.rbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost, updates = rbm.get_cost_updates(learning_rate,
persistent=None,
k=k)
# compile the theano function
fn = theano.function(
inputs=[index, theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x: train_set_x[batch_begin:batch_end]})
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns
def build_finetune_functions(self, dataset, batch_size, learning_rate,
y_type):
'''Generates a function `train` that implements one step of
finetuning, a function `validate` that computes the error on a
batch from the validation set, and a function `test` that
computes the error on a batch from the testing set
:type datasets: list of pairs of theano.tensor.TensorType
:param datasets: It is a list that contain all the datasets;
the has to contain three pairs, `train`,
`valid`, `test` in this order, where each pair
is formed of two Theano variables, one for the
datapoints, the other for the labels
:type batch_size: int
:param batch_size: size of a minibatch
:type learning_rate: float
:param learning_rate: learning rate used during finetune stage
'''
train_set_x, train_set_y = theano.shared(
dataset.phase2['train']['x']), theano.shared(
dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(
dataset.phase2['valid']['x']), theano.shared(
dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(
dataset.phase2['test']['x']), theano.shared(
dataset.phase2['test']['y'])
if not y_type == 0:
train_set_y = T.cast(train_set_y, 'int32')
valid_set_y = T.cast(valid_set_y, 'int32')
test_set_y = T.cast(test_set_y, 'int32')
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
epoch = T.scalar()
# compute the gradients with respect to the model parameters
if self.dropout:
gparams = T.grad(self.dropout_finetune_cost, self.params)
else:
gparams = T.grad(self.finetune_cost, self.params)
############################################################
gparams_mom = []
for param in self.params:
gparam_mom = theano.shared(
numpy.zeros(param.get_value(borrow=True).shape,
dtype=theano.config.floatX))
gparams_mom.append(gparam_mom)
mom = ifelse(epoch < 500,
0.5 * (1. - epoch / 500.) + 0.99 * (epoch / 500.), 0.99)
# Update the step direction using momentum
updates = {}
for gparam_mom, gparam in zip(gparams_mom, gparams):
updates[gparam_mom] = mom * gparam_mom + (1. - mom) * gparam
squared_filter_length_limit = 15.0
# ... and take a step along that direction
for param, gparam_mom in zip(self.params, gparams_mom):
stepped_param = param - learning_rate * updates[gparam_mom]
# This is a silly hack to constrain the norms of the rows of the weight
# matrices. This just checks if there are two dimensions to the
# parameter and constrains it if so... maybe this is a bit silly but it
# should work for now.
if param.get_value(borrow=True).ndim == 2:
squared_norms = T.sum(stepped_param**2, axis=1).reshape(
(stepped_param.shape[0], 1))
scale = T.clip(
T.sqrt(squared_filter_length_limit / squared_norms), 0.,
1.)
updates[param] = stepped_param * scale
else:
updates[param] = stepped_param
#########################################################
# compute list of fine-tuning updates
# updates = []
# for param, gparam in zip(self.params, gparams):
# updates.append((param, param - gparam * learning_rate))
#########################################################
output = self.dropout_finetune_cost if self.dropout else self.finetune_cost
train_fn = theano.function(
inputs=[epoch, index],
outputs=output,
updates=updates,
givens={
self.x:
train_set_x[index * batch_size:(index + 1) * batch_size],
self.y:
train_set_y[index * batch_size:(index + 1) * batch_size]
},
name='train')
test_score_i = theano.function(
[index],
self.errors,
givens={
self.x:
test_set_x[index * batch_size:(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:(index + 1) * batch_size]
},
name='test')
valid_score_i = theano.function(
[index],
self.errors,
givens={
self.x:
valid_set_x[index * batch_size:(index + 1) * batch_size],
self.y:
valid_set_y[index * batch_size:(index + 1) * batch_size]
},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1, y_type=1, gbrbm=False, dropout=False, activation_function=None):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.dropout = dropout
self.sigmoid_layers = []
self.dropout_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
rectified_linear_activation = lambda x: T.maximum(0.0, x)
# activation_function = T.nnet.sigmoid
if not activation_function:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
else:
if activation_function == 'ReLU':
print 'ReLU'
self.activation_function = rectified_linear_activation
else:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
if y_type==0:
self.y = T.matrix('y') # the labels are presented as 1D vector
else:
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
if dropout:
dropout_layer_input = _dropout_from_layer(numpy_rng, self.x, p=0.2)
else:
layer_input = self.sigmoid_layers[-1].output
if dropout:
dropout_layer_input = self.dropout_layers[-1].output
if dropout:
print 'Dropout'
dropout_layer = DropoutHiddenLayer(rng=numpy_rng,
input=dropout_layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
)
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
W=dropout_layer.W * (0.8 if i == 0 else 0.5),
b=dropout_layer.b
)
self.dropout_layers.append(dropout_layer)
self.params.extend(dropout_layer.params)
else:
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
self.params.extend(sigmoid_layer.params)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# Construct an RBM that shared weights with this layer
if dropout:
layer = dropout_layer
else:
layer = sigmoid_layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=layer.W,
hbias=layer.b,
y_type=y_type,
gbrbm=gbrbm
)
self.rbm_layers.append(rbm_layer)
if dropout:
self.dropout_output_layer = LogisticRegression(
input=self.dropout_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type)
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type,
W=self.dropout_output_layer.W * 0.5,
b=self.dropout_output_layer.b)
else:
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type)
self.get_prediction = theano.function(
inputs = [self.x],
outputs = [self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.logLayer.p_y_given_x]
)
if dropout:
self.get_prediction_dropout = theano.function(
inputs = [self.x],
outputs = [self.dropout_output_layer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.dropout_output_layer.p_y_given_x]
)
if dropout:
self.params.extend(self.dropout_output_layer.params)
else:
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
if y_type == 0:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.squared_error(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
if dropout:
self.dropout_errors = self.dropout_output_layer.errors(self.y)
self.errors = self.logLayer.errors(self.y)
class DBN(object):
"""Deep Belief Network
A deep belief network is obtained by stacking several RBMs on top of each
other. The hidden layer of the RBM at layer `i` becomes the input of the
RBM at layer `i+1`. The first layer RBM gets as input the input of the
network, and the hidden layer of the last RBM represents the output. When
used for classification, the DBN is treated as a MLP, by adding a logistic
regression layer on top.
"""
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1, y_type=1, gbrbm=False, dropout=False, activation_function=None):
"""This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the DBN
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
"""
self.dropout = dropout
self.sigmoid_layers = []
self.dropout_layers = []
self.rbm_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
rectified_linear_activation = lambda x: T.maximum(0.0, x)
# activation_function = T.nnet.sigmoid
if not activation_function:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
else:
if activation_function == 'ReLU':
print 'ReLU'
self.activation_function = rectified_linear_activation
else:
print 'Sigmoid'
self.activation_function = T.nnet.sigmoid
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
if y_type==0:
self.y = T.matrix('y') # the labels are presented as 1D vector
else:
self.y = T.ivector('y') # the labels are presented as 1D vector
# of [int] labels
# The DBN is an MLP, for which all weights of intermediate
# layers are shared with a different RBM. We will first
# construct the DBN as a deep multilayer perceptron, and when
# constructing each sigmoidal layer we also construct an RBM
# that shares weights with that layer. During pretraining we
# will train these RBMs (which will lead to chainging the
# weights of the MLP as well) During finetuning we will finish
# training the DBN by doing stochastic gradient descent on the
# MLP.
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden
# units of the layer below or the input size if we are on
# the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the
# hidden layer below or the input of the DBN if you are on
# the first layer
if i == 0:
layer_input = self.x
if dropout:
dropout_layer_input = _dropout_from_layer(numpy_rng, self.x, p=0.2)
else:
layer_input = self.sigmoid_layers[-1].output
if dropout:
dropout_layer_input = self.dropout_layers[-1].output
if dropout:
print 'Dropout'
dropout_layer = DropoutHiddenLayer(rng=numpy_rng,
input=dropout_layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
)
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=self.activation_function,
W=dropout_layer.W * (0.8 if i == 0 else 0.5),
b=dropout_layer.b
)
self.dropout_layers.append(dropout_layer)
self.params.extend(dropout_layer.params)
else:
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
self.params.extend(sigmoid_layer.params)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# Construct an RBM that shared weights with this layer
if dropout:
layer = dropout_layer
else:
layer = sigmoid_layer
rbm_layer = RBM(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=layer.W,
hbias=layer.b,
y_type=y_type,
gbrbm=gbrbm
)
self.rbm_layers.append(rbm_layer)
if dropout:
self.dropout_output_layer = LogisticRegression(
input=self.dropout_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type)
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type,
W=self.dropout_output_layer.W * 0.5,
b=self.dropout_output_layer.b)
else:
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1],
n_out=n_outs, y_type=y_type)
self.get_prediction = theano.function(
inputs = [self.x],
outputs = [self.logLayer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.logLayer.p_y_given_x]
)
if dropout:
self.get_prediction_dropout = theano.function(
inputs = [self.x],
outputs = [self.dropout_output_layer.y_pred]
)
self.get_py = theano.function(
inputs = [self.x],
outputs = [self.dropout_output_layer.p_y_given_x]
)
if dropout:
self.params.extend(self.dropout_output_layer.params)
else:
self.params.extend(self.logLayer.params)
# compute the cost for second phase of training, defined as the
# negative log likelihood of the logistic regression (output) layer
#self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
if y_type == 0:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.squared_error(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
else:
if dropout:
self.dropout_finetune_cost = self.dropout_output_layer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
if dropout:
self.dropout_errors = self.dropout_output_layer.errors(self.y)
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size, k):
'''Generates a list of functions, for performing one step of
gradient descent at a given layer. The function will require
as input the minibatch index, and to train an RBM you just
need to iterate, calling the corresponding function on all
minibatch indexes.
:type train_set_x: theano.tensor.TensorType
:param train_set_x: Shared var. that contains all datapoints used
for training the RBM
:type batch_size: int
:param batch_size: size of a [mini]batch
:param k: number of Gibbs steps to do in CD-k / PCD-k
'''
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
for rbm in self.rbm_layers:
# get the cost and the updates list
# using CD-k here (persisent=None) for training each RBM.
# TODO: change cost function to reconstruction error
cost, updates = rbm.get_cost_updates(learning_rate,
persistent=None, k=k)
# compile the theano function
fn = theano.function(inputs=[index,
theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x:
train_set_x[batch_begin:batch_end]})
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns
def build_finetune_functions(self, dataset, batch_size, learning_rate, y_type):
'''Generates a function `train` that implements one step of
finetuning, a function `validate` that computes the error on a
batch from the validation set, and a function `test` that
computes the error on a batch from the testing set
:type datasets: list of pairs of theano.tensor.TensorType
:param datasets: It is a list that contain all the datasets;
the has to contain three pairs, `train`,
`valid`, `test` in this order, where each pair
is formed of two Theano variables, one for the
datapoints, the other for the labels
:type batch_size: int
:param batch_size: size of a minibatch
:type learning_rate: float
:param learning_rate: learning rate used during finetune stage
'''
train_set_x, train_set_y = theano.shared(dataset.phase2['train']['x']), theano.shared(dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(dataset.phase2['valid']['x']), theano.shared(dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(dataset.phase2['test']['x']), theano.shared(dataset.phase2['test']['y'])
if not y_type==0:
train_set_y = T.cast(train_set_y,'int32')
valid_set_y = T.cast(valid_set_y,'int32')
test_set_y = T.cast(test_set_y,'int32')
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
epoch = T.scalar()
# compute the gradients with respect to the model parameters
if self.dropout:
gparams = T.grad(self.dropout_finetune_cost, self.params)
else:
gparams = T.grad(self.finetune_cost, self.params)
############################################################
gparams_mom = []
for param in self.params:
gparam_mom = theano.shared(numpy.zeros(param.get_value(borrow=True).shape, dtype=theano.config.floatX))
gparams_mom.append(gparam_mom)
mom = ifelse(epoch < 500,
0.5*(1. - epoch/500.) + 0.99*(epoch/500.),
0.99)
# Update the step direction using momentum
updates = {}
for gparam_mom, gparam in zip(gparams_mom, gparams):
updates[gparam_mom] = mom * gparam_mom + (1. - mom) * gparam
squared_filter_length_limit = 15.0
# ... and take a step along that direction
for param, gparam_mom in zip(self.params, gparams_mom):
stepped_param = param - learning_rate * updates[gparam_mom]
# This is a silly hack to constrain the norms of the rows of the weight
# matrices. This just checks if there are two dimensions to the
# parameter and constrains it if so... maybe this is a bit silly but it
# should work for now.
if param.get_value(borrow=True).ndim == 2:
squared_norms = T.sum(stepped_param**2, axis=1).reshape((stepped_param.shape[0],1))
scale = T.clip(T.sqrt(squared_filter_length_limit / squared_norms), 0., 1.)
updates[param] = stepped_param * scale
else:
updates[param] = stepped_param
#########################################################
# compute list of fine-tuning updates
# updates = []
# for param, gparam in zip(self.params, gparams):
# updates.append((param, param - gparam * learning_rate))
#########################################################
output = self.dropout_finetune_cost if self.dropout else self.finetune_cost
train_fn = theano.function(inputs=[epoch, index],
outputs=output,
updates=updates,
givens={self.x: train_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: train_set_y[index * batch_size:
(index + 1) * batch_size]},
name='train')
test_score_i = theano.function([index], self.errors,
givens={self.x: test_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:
(index + 1) * batch_size]},
name='test')
valid_score_i = theano.function([index], self.errors,
givens={self.x: valid_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: valid_set_y[index * batch_size:
(index + 1) * batch_size]},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1,
corruption_levels=[0.1, 0.1]):
""" This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the sdA
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
:type corruption_levels: list of float
:param corruption_levels: amount of corruption to use for each
layer
"""
self.sigmoid_layers = []
self.dA_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.matrix('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
self.params.extend(sigmoid_layer.params)
# Construct a denoising autoencoder that shared weights with this
# layer
dA_layer = dA(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
bhid=sigmoid_layer.b)
self.dA_layers.append(dA_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1], n_out=n_outs)
self.get_prediction = theano.function(
inputs=[self.x],
outputs=[self.logLayer.y_pred]
)
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
# self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
class SdA(object):
"""Stacked denoising auto-encoder class (SdA)
A stacked denoising autoencoder model is obtained by stacking several
dAs. The hidden layer of the dA at layer `i` becomes the input of
the dA at layer `i+1`. The first layer dA gets as input the input of
the SdA, and the hidden layer of the last dA represents the output.
Note that after pretraining, the SdA is dealt with as a normal MLP,
the dAs are only used to initialize the weights.
"""
def __init__(self, numpy_rng, theano_rng=None, n_ins=784,
hidden_layers_sizes=[500, 500], n_outs=1,
corruption_levels=[0.1, 0.1]):
""" This class is made to support a variable number of layers.
:type numpy_rng: numpy.random.RandomState
:param numpy_rng: numpy random number generator used to draw initial
weights
:type theano_rng: theano.tensor.shared_randomstreams.RandomStreams
:param theano_rng: Theano random generator; if None is given one is
generated based on a seed drawn from `rng`
:type n_ins: int
:param n_ins: dimension of the input to the sdA
:type n_layers_sizes: list of ints
:param n_layers_sizes: intermediate layers size, must contain
at least one value
:type n_outs: int
:param n_outs: dimension of the output of the network
:type corruption_levels: list of float
:param corruption_levels: amount of corruption to use for each
layer
"""
self.sigmoid_layers = []
self.dA_layers = []
self.params = []
self.n_layers = len(hidden_layers_sizes)
assert self.n_layers > 0
if not theano_rng:
theano_rng = RandomStreams(numpy_rng.randint(2 ** 30))
# allocate symbolic variables for the data
self.x = T.matrix('x') # the data is presented as rasterized images
self.y = T.matrix('y') # the labels are presented as 1D vector of
# [int] labels
# The SdA is an MLP, for which all weights of intermediate layers
# are shared with a different denoising autoencoders
# We will first construct the SdA as a deep multilayer perceptron,
# and when constructing each sigmoidal layer we also construct a
# denoising autoencoder that shares weights with that layer
# During pretraining we will train these autoencoders (which will
# lead to chainging the weights of the MLP as well)
# During finetunining we will finish training the SdA by doing
# stochastich gradient descent on the MLP
for i in xrange(self.n_layers):
# construct the sigmoidal layer
# the size of the input is either the number of hidden units of
# the layer below or the input size if we are on the first layer
if i == 0:
input_size = n_ins
else:
input_size = hidden_layers_sizes[i - 1]
# the input to this layer is either the activation of the hidden
# layer below or the input of the SdA if you are on the first
# layer
if i == 0:
layer_input = self.x
else:
layer_input = self.sigmoid_layers[-1].output
sigmoid_layer = HiddenLayer(rng=numpy_rng,
input=layer_input,
n_in=input_size,
n_out=hidden_layers_sizes[i],
activation=T.nnet.sigmoid)
# add the layer to our list of layers
self.sigmoid_layers.append(sigmoid_layer)
# its arguably a philosophical question...
# but we are going to only declare that the parameters of the
# sigmoid_layers are parameters of the StackedDAA
# the visible biases in the dA are parameters of those
# dA, but not the SdA
self.params.extend(sigmoid_layer.params)
# Construct a denoising autoencoder that shared weights with this
# layer
dA_layer = dA(numpy_rng=numpy_rng,
theano_rng=theano_rng,
input=layer_input,
n_visible=input_size,
n_hidden=hidden_layers_sizes[i],
W=sigmoid_layer.W,
bhid=sigmoid_layer.b)
self.dA_layers.append(dA_layer)
# We now need to add a logistic layer on top of the MLP
self.logLayer = LogisticRegression(
input=self.sigmoid_layers[-1].output,
n_in=hidden_layers_sizes[-1], n_out=n_outs)
self.get_prediction = theano.function(
inputs=[self.x],
outputs=[self.logLayer.y_pred]
)
self.params.extend(self.logLayer.params)
# construct a function that implements one step of finetunining
# compute the cost for second phase of training,
# defined as the negative log likelihood
# self.finetune_cost = self.logLayer.negative_log_likelihood(self.y)
self.finetune_cost = self.logLayer.squared_error(self.y)
# compute the gradients with respect to the model parameters
# symbolic variable that points to the number of errors made on the
# minibatch given by self.x and self.y
self.errors = self.logLayer.errors(self.y)
def pretraining_functions(self, train_set_x, batch_size):
''' Generates a list of functions, each of them implementing one
step in trainnig the dA corresponding to the layer with same index.
The function will require as input the minibatch index, and to train
a dA you just need to iterate, calling the corresponding function on
all minibatch indexes.
:type train_set_x: theano.tensor.TensorType
:param train_set_x: Shared variable that contains all datapoints used
for training the dA
:type batch_size: int
:param batch_size: size of a [mini]batch
:type learning_rate: float
:param learning_rate: learning rate used during training for any of
the dA layers
'''
# index to a [mini]batch
index = T.lscalar('index') # index to a minibatch
corruption_level = T.scalar('corruption') # % of corruption to use
learning_rate = T.scalar('lr') # learning rate to use
# number of batches
n_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size
# begining of a batch, given `index`
batch_begin = index * batch_size
# ending of a batch given `index`
batch_end = batch_begin + batch_size
pretrain_fns = []
for dA in self.dA_layers:
# get the cost and the updates list
cost, updates = dA.get_cost_updates(corruption_level,
learning_rate)
# compile the theano function
fn = theano.function(inputs=[index,
theano.Param(corruption_level, default=0.2),
theano.Param(learning_rate, default=0.1)],
outputs=cost,
updates=updates,
givens={self.x: train_set_x[batch_begin:
batch_end]})
# append `fn` to the list of functions
pretrain_fns.append(fn)
return pretrain_fns
def build_finetune_functions(self, dataset, batch_size, learning_rate):
'''Generates a function `train` that implements one step of
finetuning, a function `validate` that computes the error on
a batch from the validation set, and a function `test` that
computes the error on a batch from the testing set
:type datasets: list of pairs of theano.tensor.TensorType
:param datasets: It is a list that contain all the datasets;
the has to contain three pairs, `train`,
`valid`, `test` in this order, where each pair
is formed of two Theano variables, one for the
datapoints, the other for the labels
:type batch_size: int
:param batch_size: size of a minibatch
:type learning_rate: float
:param learning_rate: learning rate used during finetune stage
'''
train_set_x, train_set_y = theano.shared(dataset.phase2['train']['x']), theano.shared(dataset.phase2['train']['y'])
valid_set_x, valid_set_y = theano.shared(dataset.phase2['valid']['x']), theano.shared(dataset.phase2['valid']['y'])
test_set_x, test_set_y = theano.shared(dataset.phase2['test']['x']), theano.shared(dataset.phase2['test']['y'])
# compute number of minibatches for training, validation and testing
n_valid_batches = valid_set_x.get_value(borrow=True).shape[0]
n_valid_batches /= batch_size
n_test_batches = test_set_x.get_value(borrow=True).shape[0]
n_test_batches /= batch_size
index = T.lscalar('index') # index to a [mini]batch
# compute the gradients with respect to the model parameters
gparams = T.grad(self.finetune_cost, self.params)
# compute list of fine-tuning updates
updates = []
for param, gparam in zip(self.params, gparams):
updates.append((param, param - gparam * learning_rate))
train_fn = theano.function(inputs=[index],
outputs=self.finetune_cost,
updates=updates,
givens={
self.x: train_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: train_set_y[index * batch_size:
(index + 1) * batch_size]},
name='train')
test_score_i = theano.function([index], self.errors,
givens={
self.x: test_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: test_set_y[index * batch_size:
(index + 1) * batch_size]},
name='test')
valid_score_i = theano.function([index], self.errors,
givens={
self.x: valid_set_x[index * batch_size:
(index + 1) * batch_size],
self.y: valid_set_y[index * batch_size:
(index + 1) * batch_size]},
name='valid')
# Create a function that scans the entire validation set
def valid_score():
return [valid_score_i(i) for i in xrange(n_valid_batches)]
# Create a function that scans the entire test set
def test_score():
return [test_score_i(i) for i in xrange(n_test_batches)]
return train_fn, valid_score, test_score