def _train_fprop(self, state_below):
conv_out = conv2d_fft(state_below,
self.W,
border_mode=self.border_mode,
image_shape=self.image_shape,
pad_last_dim=self.pad_last_dim)
return conv_out + self.b.dimshuffle('x', 0, 'x', 'x')
def _train_fprop(self, state_below):
conv_out = conv2d_fft(
state_below,
self.W,
border_mode=self.border_mode,
image_shape=self.image_shape,
pad_last_dim=self.pad_last_dim,
)
return conv_out + self.b.dimshuffle("x", 0, "x", "x")
def output_func(self, input):
return conv2d_fft(input, self.W, border_mode='valid',
filter_shape=self.filter_shape,
image_shape=self.input_shape)
def output_func(self, input):
return conv2d_fft(input,
self.W,
border_mode='valid',
filter_shape=self.filter_shape,
image_shape=self.input_shape)
def __init__(self,
rng,
input,
subsample,
filter_shape,
image_shape,
poolsize=(2, 2)):
"""
Allocate a LeNetConvPoolLayer with shared variable internal parameters.
:type rng: numpy.random.RandomState
:param rng: a random number generator used to initialize weights
:type input: theano.tensor.dtensor4
:param input: symbolic image tensor, of shape image_shape
:type filter_shape: tuple or list of length 4
:param filter_shape: (number of filters, num input feature maps,
filter height, filter width)
:type image_shape: tuple or list of length 4
:param image_shape: (batch size, num input feature maps,
image height, image width)
:type poolsize: tuple or list of length 2
:param poolsize: the downsampling (pooling) factor (#rows, #cols)
"""
assert image_shape[1] == filter_shape[1]
self.input = input
# there are "num input feature maps * filter height * filter width"
# inputs to each hidden unit
fan_in = numpy.prod(filter_shape[1:])
# each unit in the lower layer receives a gradient from:
# "num output feature maps * filter height * filter width" /
# pooling size
fan_out = (filter_shape[0] * numpy.prod(filter_shape[2:]) /
numpy.prod(poolsize))
# initialize weights with random weights
W_bound = numpy.sqrt(6. / (fan_in + fan_out))
self.W = theano.tensor._shared(numpy.asarray(
rng.uniform(low=-W_bound, high=W_bound, size=filter_shape),
dtype=theano.config.floatX),
borrow=True)
# the bias is a 1D tensor -- one bias per output feature map
b_values = numpy.zeros((filter_shape[0], ), dtype=theano.config.floatX)
self.b = theano.tensor._shared(value=b_values, borrow=True)
fft_filters = False
# convolve input feature maps with filters
if fft_filters != True:
conv_out = conv.conv2d(input=input,
filters=self.W,
filter_shape=filter_shape,
subsample=subsample,
image_shape=image_shape)
elif fft_filters is True:
conv_out = conv2d_fft(
input=theano.sandbox.cuda.as_cuda_ndarray_variable(input),
filters=self.W,
filter_shape=filter_shape,
image_shape=image_shape)
# downsample each feature map individually, using maxpooling
pooled_out = downsample.max_pool_2d(input=conv_out,
ds=poolsize,
ignore_border=True)
# add the bias term. Since the bias is a vector (1D array), we first
# reshape it to a tensor of shape (1, n_filters, 1, 1). Each bias will
# thus be broadcasted across mini-batches and feature map
# width & height
def rectify(X):
return T.maximum(X, 0.)
self.output = rectify(pooled_out + self.b.dimshuffle('x', 0, 'x', 'x'))
# store parameters of this layer
self.params = [self.W, self.b]
