# Prepare Theano variables for inputs and targets
input = T.tensor4('inputs')
target = T.matrix('targets')
LR = T.scalar('LR', dtype=theano.config.floatX)
mlp = lasagne.layers.InputLayer(shape=(None, 1, 28, 28), input_var=input)
mlp = lasagne.layers.DropoutLayer(mlp, p=dropout_in)
for k in range(n_hidden_layers):
mlp = binary_connect.DenseLayer(
mlp,
binary=binary,
stochastic=stochastic,
H=H,
nonlinearity=lasagne.nonlinearities.identity,
num_units=num_units)
mlp = batch_norm.BatchNormLayer(
mlp,
epsilon=epsilon,
alpha=alpha,
nonlinearity=lasagne.nonlinearities.rectify)
mlp = lasagne.layers.DropoutLayer(mlp, p=dropout_hidden)
mlp = binary_connect.DenseLayer(
mlp,
binary=binary,
cnn = lasagne.layers.MaxPool2DLayer(cnn, pool_size=(2, 2))
cnn = batch_norm.BatchNormLayer(
cnn,
epsilon=epsilon,
alpha=alpha,
nonlinearity=lasagne.nonlinearities.rectify)
# print(cnn.output_shape)
# 1024FP-1024FP-10FP
cnn = binary_connect.DenseLayer(
cnn,
binary=binary,
stochastic=stochastic,
H=H,
W_LR_scale=W_LR_scale,
nonlinearity=lasagne.nonlinearities.identity,
num_units=1024)
cnn = batch_norm.BatchNormLayer(
cnn,
epsilon=epsilon,
alpha=alpha,
nonlinearity=lasagne.nonlinearities.rectify)
cnn = binary_connect.DenseLayer(
cnn,
binary=binary,
stochastic=stochastic,
H=H,