def mapping(self, x):
# Transform unconstrained parameters to lie on simplex.
pi = Variable("pi", [self.num_factors, self.K - 1])
pi_const = tf.sigmoid(pi)
pi_const = tf.concat(
1, [pi_const,
tf.expand_dims(1.0 - tf.reduce_sum(pi_const), 0)])
return [pi_const]
def mapping(self, x):
p = Variable("p", [self.num_params])
# Constrain parameters to lie on simplex.
p_const = tf.sigmoid(p)
if self.num_vars > 1:
p_const = tf.concat(
0, [p_const,
tf.expand_dims(1.0 - tf.reduce_sum(p_const), 0)])
return [p_const]
def mapping(self, x):
# Transform a real (K-1)-vector to K-dimensional simplex.
pi = Variable("pi", [self.num_factors, self.K - 1])
eq = -tf.log(
tf.cast(self.K - 1 - tf.range(self.K - 1), dtype=tf.float32))
z = tf.sigmoid(eq + pi)
pil = tf.concat(1, [z, tf.ones([self.num_factors, 1])])
piu = tf.concat(1, [tf.ones([self.num_factors, 1]), 1.0 - z])
# cumulative product along 1st axis
S = tf.pack([cumprod(piu_x) for piu_x in tf.unpack(piu)])
return [S * pil]
def mapping(self, x):
params = Variable("params", [self.num_vars])
return [self.transform(params)]
def mapping(self, x):
mean = Variable("mu", [self.num_vars])
stddev = Variable("sigma", [self.num_vars])
return [tf.identity(mean), tf.nn.softplus(stddev)]
def mapping(self, x):
alpha = Variable("alpha", [self.num_vars])
beta = Variable("beta", [self.num_vars])
return [tf.nn.softplus(alpha) + 1e-2, tf.nn.softplus(beta) + 1e-2]
def mapping(self, x):
alpha = Variable("dirichlet_alpha", [self.num_factors, self.K])
return [tf.nn.softplus(alpha)]
def mapping(self, x):
p = Variable("p", [self.num_params])
return [tf.sigmoid(p)]