def build_nips_network(self, input_width, input_height, output_dim,
num_frames, batch_size, batch_norm = False):
"""
Build a network consistent with the 2013 NIPS paper.
"""
from lasagne.layers import cuda_convnet
l_in = lasagne.layers.InputLayer(
shape=( None, num_frames, input_width, input_height)
)
l_conv1 = cuda_convnet.Conv2DCCLayer(
l_in,
num_filters=16,
filter_size=(8, 8),
stride=(4, 4),
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(c01b=True),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1),
dimshuffle=True
)
if batch_norm :
l_conv1 = batch_norm_layer( l_conv1 )
l_conv2 = cuda_convnet.Conv2DCCLayer(
l_conv1,
num_filters=32,
filter_size=(4, 4),
stride=(2, 2),
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(c01b=True),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1),
dimshuffle=True
)
if batch_norm :
l_conv2 = batch_norm_layer( l_conv2 )
l_hidden1 = lasagne.layers.DenseLayer(
l_conv2,
num_units=256,
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_hidden1 = batch_norm_layer( l_hidden1 )
l_out = lasagne.layers.DenseLayer(
l_hidden1,
num_units=output_dim,
nonlinearity=None,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
return l_out
def build_nips_network_dnn(self, input_width, input_height, output_dim,
num_frames, batch_size, batch_norm = False):
"""
Build a network consistent with the 2013 NIPS paper.
"""
# Import it here, in case it isn't installed.
from lasagne.layers import dnn
l_in = lasagne.layers.InputLayer(
shape=( None, num_frames, input_width, input_height)
)
l_conv1 = dnn.Conv2DDNNLayer(
l_in,
num_filters=16,
filter_size=(8, 8),
stride=(4, 4),
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_conv1 = batch_norm_layer( l_conv1 )
l_conv2 = dnn.Conv2DDNNLayer(
l_conv1,
num_filters=32,
filter_size=(4, 4),
stride=(2, 2),
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_conv2 = batch_norm_layer( l_conv2 )
l_hidden1 = lasagne.layers.DenseLayer(
l_conv2,
num_units=256,
nonlinearity=lasagne.nonlinearities.rectify,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_hidden1 = batch_norm_layer( l_hidden1 )
l_out = lasagne.layers.DenseLayer(
l_hidden1,
num_units=output_dim,
nonlinearity=None,
#W=lasagne.init.HeUniform(),
W=lasagne.init.Normal(.01),
b=lasagne.init.Constant(.1)
)
return l_out
def build_nature_network_dnn(self, input_width, input_height, output_dim,
num_frames, batch_size, batch_norm = False):
"""
Build a large network consistent with the DeepMind Nature paper.
"""
from lasagne.layers import dnn
l_in = lasagne.layers.InputLayer(
shape=( None, num_frames, input_width, input_height)
)
l_conv1 = dnn.Conv2DDNNLayer(
l_in,
num_filters=32,
filter_size=(8, 8),
stride=(4, 4),
nonlinearity=lasagne.nonlinearities.rectify,
W=lasagne.init.HeUniform(),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_conv1 = batch_norm_layer( l_conv1 )
l_conv2 = dnn.Conv2DDNNLayer(
l_conv1,
num_filters=64,
filter_size=(4, 4),
stride=(2, 2),
nonlinearity=lasagne.nonlinearities.rectify,
W=lasagne.init.HeUniform(),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_conv2 = batch_norm_layer( l_conv2 )
l_conv3 = dnn.Conv2DDNNLayer(
l_conv2,
num_filters=64,
filter_size=(3, 3),
stride=(1, 1),
nonlinearity=lasagne.nonlinearities.rectify,
W=lasagne.init.HeUniform(),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_conv3 = batch_norm_layer( l_conv3 )
l_hidden1 = lasagne.layers.DenseLayer(
l_conv3,
num_units=512,
nonlinearity=lasagne.nonlinearities.rectify,
W=lasagne.init.HeUniform(),
b=lasagne.init.Constant(.1)
)
if batch_norm :
l_hidden1 = batch_norm_layer( l_hidden1 )
l_out = lasagne.layers.DenseLayer(
l_hidden1,
num_units=output_dim,
nonlinearity=None,
W=lasagne.init.HeUniform(),
b=lasagne.init.Constant(.1)
)
return l_out
