lunny/keras
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Switch convolutional layers to new regularization

Browse Source
This commit is contained in:
fchollet 2015-06-27 15:43:49 -07:00
parent f04fdec9e6
commit 8ef90a03b3
1 changed files with 24 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
keras/layers/convolutional.py
View File

@ -14,7 +14,7 @@ class Convolution1D(Layer):
def __init__(self, nb_filter, stack_size, filter_length,
init='uniform', activation='linear', weights=None,
image_shape=None, border_mode='valid', subsample_length=1,
W_regularizer=None, b_regularizer=None, W_constraint=None, b_constraint=None):
W_regularizer=None, b_regularizer=None, activity_regularizer=None, W_constraint=None, b_constraint=None):
nb_row = 1
nb_col = filter_length
@ -36,7 +36,17 @@ class Convolution1D(Layer):
self.params = [self.W, self.b]
self.regularizers = [W_regularizer, b_regularizer]
self.regularizers = []
if W_regularizer:
W_regularizer.set_param(self.W)
self.regularizers.append(W_regularizer)
if b_regularizer:
b_regularizer.set_param(self.b)
self.regularizers.append(b_regularizer)
if activity_regularizer:
activity_regularizer.set_layer(self)
self.regularizers.append(activity_regularizer)
self.constraints = [W_constraint, b_constraint]
if weights is not None:
@ -87,7 +97,7 @@ class Convolution2D(Layer):
def __init__(self, nb_filter, stack_size, nb_row, nb_col,
init='glorot_uniform', activation='linear', weights=None,
image_shape=None, border_mode='valid', subsample=(1,1),
W_regularizer=None, b_regularizer=None, W_constraint=None, b_constraint=None):
W_regularizer=None, b_regularizer=None, activity_regularizer=None, W_constraint=None, b_constraint=None):
super(Convolution2D,self).__init__()
self.init = initializations.get(init)
@ -107,7 +117,17 @@ class Convolution2D(Layer):
self.params = [self.W, self.b]
self.regularizers = [W_regularizer, b_regularizer]
self.regularizers = []
if W_regularizer:
W_regularizer.set_param(self.W)
self.regularizers.append(W_regularizer)
if b_regularizer:
b_regularizer.set_param(self.b)
self.regularizers.append(b_regularizer)
if activity_regularizer:
activity_regularizer.set_layer(self)
self.regularizers.append(activity_regularizer)
self.constraints = [W_constraint, b_constraint]
if weights is not None: