wassname/keras-contrib
1
0
Fork 0
mirror of https://github.com/wassname/keras-contrib.git synced 2026-08-02 12:50:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update normalization.py

Browse Source
This commit is contained in:
Michael Oliver
2017-08-16 17:41:04 -07:00
committed by GitHub
parent 5eaa736fee
commit 7c19245824
1 changed files with 3 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
keras_contrib/layers/normalization.py
+3 -5
View File
@@ -318,14 +318,12 @@ class BatchRenormalization(Layer):
K.moving_average_update(self.running_variance, std_batch ** 2, self.momentum)], inputs)
# update r_max and d_max
t_val = K.get_value(self.t)
r_val = self.r_max_value / (1 + (self.r_max_value - 1) * K.exp(-t_val))
d_val = self.d_max_value / (1 + ((self.d_max_value / 1e-3) - 1) * K.exp(-(2 * t_val)))
t_val += float(self.t_delta)
r_val = self.r_max_value / (1 + (self.r_max_value - 1) * K.exp(-self.t))
d_val = self.d_max_value / (1 + ((self.d_max_value / 1e-3) - 1) * K.exp(-(2 * self.t)))
self.add_update([K.update(self.r_max, r_val),
K.update(self.d_max, d_val),
K.update(self.t, t_val)], inputs)
K.update_add(self.t, K.variable(np.array([self.t_delta])))], x)
if training in {0, False}:
return x_normed