def make_summaries(self, env):
with self.graph.as_default():
if env.is_image():
idx = T.random_uniform([], minval = 0, maxval = self.horizon - 1, dtype = T.int32)
env.make_summary(self.q_O.get_parameters('regular')[0:1][:, idx], "reconstruction")
env.make_summary(self.O[0:1][:, idx], "truth")
self.summary = T.core.summary.merge_all()
def xavier(shape, constant=1):
""" Xavier initialization of network weights"""
fan_in, fan_out = get_fans(shape)
low = -constant*np.sqrt(6.0/(fan_in + fan_out))
high = constant*np.sqrt(6.0/(fan_in + fan_out))
return T.random_uniform(shape,
minval=low, maxval=high,
dtype=T.floatx())
def uniform(shape, scale=0.05):
return T.random_uniform(shape, minval=-scale, maxval=scale)
def _sample(self, num_samples):
shape = self.shape()
sample_shape = T.concat([[num_samples], shape], 0)
random_sample = T.random_uniform(sample_shape)
m, b = Stats.X(self.m), Stats.X(self.b)
return m[None] - b[None] * T.log(-T.log(random_sample))