def _initialize_theta(self):
filter_shape = self.filter_shape
image_shape = self.image_shape
poolsize = self.poolsize
conv_in = np.prod(filter_shape[1:])
conv_out = filter_shape[0] * np.prod(filter_shape[2:])
pool_out = conv_out / poolsize**2
conv_map_size = image_shape[-1] - filter_shape[-1] + 1
assert conv_map_size > 0
pool_map_size = int(conv_map_size / poolsize)
assert pool_map_size > 0
self.conv_w = shared(
nn_random_paramters(conv_in, conv_out, shape=filter_shape))
self.conv_b = shared(
nn_random_paramters(conv_in, conv_out,
shape=(filter_shape[0], 1, 1)))
self.pool_w = shared(
nn_random_paramters(conv_out, pool_out,
shape=(filter_shape[0], 1, 1)))
self.pool_b = shared(
nn_random_paramters(conv_out, pool_out,
shape=(filter_shape[0], 1, 1)))
self.output_shape = (image_shape[0], filter_shape[0],
pool_map_size, pool_map_size)
return [self.conv_w, self.conv_b, self.pool_w, self.pool_b]
def __init__(self, features, values):
assert isinstance(features, np.ndarray) and features.ndim >= 1
assert isinstance(values, np.ndarray)
assert features.shape[0] == values.shape[0]
self.size = values.shape[0]
origin_X = shared(features)
origin_y = shared(values.flatten())
self.origin = (origin_X, origin_y)
self.X = T.cast(origin_X, features.dtype.name)
self.y = T.cast(origin_y, values.dtype.name)
def __init__(self, input_size, hidden_layers_size, output_size,
output_layer=LogisticRegression, lamda=0):
"""
Parameters:
------
inpute_size: int
The number of input layer units.
hidden_layers_size: list
A list of numbers of units in each hidden layer.
output_size: int
The number of output layer units.
output_layer: object
lamda: float
Parameter used for regularization. Set 0 to disable regularize.
"""
assert isinstance(hidden_layers_size, list) and \
len(hidden_layers_size) > 0
self.lamda = shared(lamda, name='lamda')
self.layers_size = [input_size, output_size]
self.layers_size[1:-1] = hidden_layers_size
self.hidden_layers = [
HiddenLayer(self.layers_size[i-1], self.layers_size[i],
self.lamda, activation=tanh)
for i in range(1, len(self.layers_size)-1)
]
self.output_layer = output_layer(
self.layers_size[-2], self.layers_size[-1], self.lamda.get_value())
self.layers = self.hidden_layers + [self.output_layer]
def __init__(self, input_size, conv_layers, hidden_layers, output_size,
output_layer=LogisticRegression, conv_filter_size=5,
conv_pool_size=2, lamda=0):
"""
Parameters:
------
inpute_size: int
The width of input image.(assume that input image is square)
conv_layers: list
A list of numbers of feature maps in each convlution layer.
hidden_layers: list
A list of numbers of units in each hidden layer
of fully connected mlp.
output_size: int
The number of output layer units.
output_layer: object
conv_filter_size: int
The convlution factor.
conv_pool_size: int
The downsampling (pooling) factor.
lamda: float
Parameter used for regularization. Set 0 to disable regularize.
"""
self.lamda = shared(lamda, name='lamda')
image_shape = (None, 1, input_size, input_size)
self.conv_layers = []
for c in conv_layers:
filter_shape = (
c, image_shape[1], conv_filter_size, conv_filter_size
)
cp = LeNetConvPoolLayer(
filter_shape, image_shape, lamda, conv_pool_size)
image_shape = cp.output_shape
self.conv_layers.append(cp)
self.mlp = MLP(np.prod(image_shape[1:]), hidden_layers, output_size,
output_layer=output_layer, lamda=self.lamda.get_value())
self.layers = self.conv_layers + [self.mlp]
def __init__(self, n_in, n_out, lamda):
"""
n_in: int
Number of input units, the dimension of the space
in which the input lie.
n_out: int
Number of output units, the dimension of the space
in which the output lie.
lamda: theano like
Parameter used for regularization. Set 0 to disable regularize.
"""
self.n_in = n_in
self.n_out = n_out
self.lamda = lamda
self._theta = shared(self._initialize_theta())
self.w = self._theta[1:, :]
self.b = self._theta[0, :]
def __init__(self, features, labels, lamda=0):
lamda = shared(lamda, name='lamda')
super(LogisticRegression, self).__init__(features, labels, lamda)
