def log_conditional_prob(self, x):
# E_{Q(h|x)} log Q(h|x)
h_mu, h_logvar = self.inference(x)
h = sample_from_gaussian(h_mu, h_logvar)
logp_h_given_x = tf.reduce_sum(
-0.5 * np.log(2 * np.pi) - 0.5 * h_logvar -
0.5 * tf.square(h - h_mu) / tf.exp(h_logvar),
axis=-1)
return logp_h_given_x, h
def get_independent_means(self,
num_samples,
burn_in_steps=100000,
random=True):
v = tf.zeros([num_samples, self.vis_dim], dtype=tf.float32)
if random:
v = sample_from_gaussian(v, 2. * tf.log(self.sigma)) + self.vis_b
for i in xrange(burn_in_steps):
_, v = self.gibbs_vhv(v)
h_1 = sample_from_bernoulli(self.vis2hid(v))
v_1 = self.hid2vis(h_1)
return v_1.numpy()
def get_independent_samples(self,
num_samples,
burn_in_steps=100000,
random=True,
initial_v=None):
if initial_v is not None:
v = initial_v
else:
v = tf.zeros([num_samples, self.vis_dim],
dtype=tf.float32) + self.vis_b
if random:
v = sample_from_gaussian(v, 2. * tf.log(self.sigma))
for i in xrange(burn_in_steps):
_, v = self.gibbs_vhv(v)
return v.numpy()
def get_samples_single_chain(self,
num_samples,
adjacent_samples=10,
steps_between_samples=1000,
burn_in_steps=100000,
random=True):
assert num_samples % adjacent_samples == 0
v = tf.zeros([1, self.vis_dim], dtype=tf.float32) + self.vis_b
if random:
v = sample_from_gaussian(v, 2. * tf.log(self.sigma))
for i in xrange(burn_in_steps):
_, v = self.gibbs_vhv(v)
sample_list = []
for i in xrange(num_samples / adjacent_samples):
for j in xrange(adjacent_samples):
_, v = self.gibbs_vhv(v)
sample_list.append(v.numpy())
for i in xrange(steps_between_samples):
_, v = self.gibbs_vhv(v)
return np.vstack(sample_list)
def get_h(self, x):
h_mu, h_logvar = self.inference(x)
h = sample_from_gaussian(h_mu, h_logvar)
return h
def gibbs_hvh(self, h_0):
v_1 = sample_from_gaussian(self.hid2vis(h_0), 2 * tf.log(self.sigma))
v_1 = tf.stop_gradient(v_1)
h_1 = sample_from_bernoulli(self.vis2hid(v_1))
return v_1, h_1
def gibbs_vhv(self, v_0):
h_1 = sample_from_bernoulli(self.vis2hid(v_0))
v_1 = sample_from_gaussian(self.hid2vis(h_1), 2 * tf.log(self.sigma))
v_1 = tf.stop_gradient(v_1)
return h_1, v_1