def call(self, u_vecs):
if self.share_weights:
u_hat_vecs = K.conv1d(u_vecs, self.W) #(400，2*16)
else:
u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])
batch_size = K.shape(u_vecs)[0] #none
input_num_capsule = K.shape(u_vecs)[1] #400
u_hat_vecs = K.reshape(
u_hat_vecs, (batch_size, input_num_capsule, self.num_capsule,
self.dim_capsule)) #(none,400,2,16)
u_hat_vecs = K.permute_dimensions(u_hat_vecs,
(0, 2, 1, 3)) #(none,2,400,16)
b = K.zeros_like(
u_hat_vecs[:, :, :, 0]
) # shape = [None, num_capsule, input_num_capsule](none,2,400)
for i in range(self.routings):
b = K.permute_dimensions(
b, (0, 2, 1)
) # shape = [None, input_num_capsule, num_capsule](none,400,2)
c = K.softmax(b)
c = K.permute_dimensions(c, (0, 2, 1))
b = K.permute_dimensions(b, (0, 2, 1))
outputs = self.activation(K.batch_dot(c, u_hat_vecs, [2, 2]))
if i < self.routings - 1:
b += K.batch_dot(outputs, u_hat_vecs, [2, 3])
return outputs
def routing(self, u_vecs):
if self.share_weights:
u_hat_vecs = K.conv1d(u_vecs, self.W)
else:
u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])
batch_size = K.shape(u_vecs)[0]
input_num_capsule = K.shape(u_vecs)[1]
u_hat_vecs = K.reshape(u_hat_vecs,
(batch_size, input_num_capsule,
self.num_capsule, self.dim_capsule))
u_hat_vecs = K.permute_dimensions(u_hat_vecs, (0, 2, 1, 3))
# final u_hat_vecs.shape = [None, num_capsule, input_num_capsule, dim_capsule]
b = K.zeros_like(
u_hat_vecs[:, :, :,
0]) # shape = [None, num_capsule, input_num_capsule]
for i in range(self.routings):
c = K.softmax(b)
s = K.batch_dot(c, u_hat_vecs, [2, 2])
v = self.squash(s)
if i < self.routings - 1:
b = b + K.batch_dot(v, u_hat_vecs, [2, 3])
poses = v
activations = K.sqrt(K.sum(K.square(poses), 2))
return poses, activations
def call(self, x):
height, width, in_channels = x.shape.as_list()[1:]
if self.groups is None:
if in_channels % self.groups_factor:
raise ValueError("%s %% %s" %
(in_channels, self.groups_factor))
self.groups = in_channels // self.groups_factor
channels_per_group = in_channels // self.groups
x = K.reshape(x, [-1, height, width, self.groups, channels_per_group])
x = K.permute_dimensions(x, (0, 1, 2, 4, 3)) # transpose
x = K.reshape(x, [-1, height, width, in_channels])
return x
def step(self, inputs, states):
h_tm1 = states[0]
c_tm1 = states[1]
dp_mask = states[2]
rec_dp_mask = states[3]
x_input = states[4]
# alignment model
h_att = K.repeat(h_tm1, self.timestep_dim)
att = _time_distributed_dense(x_input,
self.attention_weights,
self.attention_bias,
output_dim=K.int_shape(
self.attention_weights)[1])
attention_ = self.attention_activation(
K.dot(h_att, self.attention_recurrent_weights) + att)
attention_ = K.squeeze(
K.dot(attention_, self.attention_recurrent_bias), 2)
alpha = K.exp(attention_)
if dp_mask is not None:
alpha *= dp_mask[0]
alpha /= K.sum(alpha, axis=1, keepdims=True)
alpha_r = K.repeat(alpha, self.input_dim)
alpha_r = K.permute_dimensions(alpha_r, (0, 2, 1))
# make context vector (soft attention after Bahdanau et al.)
z_hat = x_input * alpha_r
context_sequence = z_hat
z_hat = K.sum(z_hat, axis=1)
if self.implementation == 2:
z = K.dot(inputs * dp_mask[0], self.kernel)
z += K.dot(h_tm1 * rec_dp_mask[0], self.recurrent_kernel)
z += K.dot(z_hat, self.attention_kernel)
if self.use_bias:
z = K.bias_add(z, self.bias)
z0 = z[:, :self.units]
z1 = z[:, self.units:2 * self.units]
z2 = z[:, 2 * self.units:3 * self.units]
z3 = z[:, 3 * self.units:]
i = self.recurrent_activation(z0)
f = self.recurrent_activation(z1)
c = f * c_tm1 + i * self.activation(z2)
o = self.recurrent_activation(z3)
else:
if self.implementation == 0:
x_i = inputs[:, :self.units]
x_f = inputs[:, self.units:2 * self.units]
x_c = inputs[:, 2 * self.units:3 * self.units]
x_o = inputs[:, 3 * self.units:]
elif self.implementation == 1:
x_i = K.dot(inputs * dp_mask[0], self.kernel_i) + self.bias_i
x_f = K.dot(inputs * dp_mask[1], self.kernel_f) + self.bias_f
x_c = K.dot(inputs * dp_mask[2], self.kernel_c) + self.bias_c
x_o = K.dot(inputs * dp_mask[3], self.kernel_o) + self.bias_o
else:
raise ValueError('Unknown `implementation` mode.')
i = self.recurrent_activation(
x_i + K.dot(h_tm1 * rec_dp_mask[0], self.recurrent_kernel_i) +
K.dot(z_hat, self.attention_i))
f = self.recurrent_activation(
x_f + K.dot(h_tm1 * rec_dp_mask[1], self.recurrent_kernel_f) +
K.dot(z_hat, self.attention_f))
c = f * c_tm1 + i * self.activation(
x_c + K.dot(h_tm1 * rec_dp_mask[2], self.recurrent_kernel_c) +
K.dot(z_hat, self.attention_c))
o = self.recurrent_activation(
x_o + K.dot(h_tm1 * rec_dp_mask[3], self.recurrent_kernel_o) +
K.dot(z_hat, self.attention_o))
h = o * self.activation(c)
if 0 < self.dropout + self.recurrent_dropout:
h._uses_learning_phase = True
if self.return_attention:
return context_sequence, [h, c]
else:
return h, [h, c]
