def calc_z_prior(self, y):
z_mean_prior = dgm.forwardPass(self.p_z_y_mean, y)
z_log_var_prior = dgm.forwardPass(self.p_z_y_log_var, y)
z_mean_prior, z_log_var_prior = tf.reshape(
z_mean_prior, [self.mc_samples, -1, self.n_z]), tf.reshape(
z_log_var_prior, [self.mc_samples, -1, self.n_z])
return z_mean_prior, z_log_var_prior
def sample_z(self, x, y, n_samples=None):
if n_samples == None:
n_samples = self.mc_samples
l_qz_in = tf.concat([x, y], axis=-1)
z_mean = dgm.forwardPass(self.q_z_xy_mean, l_qz_in)
z_log_var = dgm.forwardPass(self.q_z_xy_log_var, l_qz_in)
return z_mean, z_log_var, dgm.sampleNormal(z_mean, z_log_var,
n_samples)
def sample_a(self, x, n_samples=None):
if n_samples == None:
n_samples = self.mc_samples
l_qa_in = x
a_mean = dgm.forwardPass(self.q_a_x_mean, l_qa_in)
a_log_var = dgm.forwardPass(self.q_a_x_log_var, l_qa_in)
return a_mean, a_log_var, dgm.sampleNormal(a_mean, a_log_var,
n_samples)
def sample_z(self, x, y, a, n_samples=None):
if n_samples == None:
n_samples = 1
l_qz_in = tf.reshape(tf.concat([x, y, a], axis=-1),
[-1, self.n_x + self.n_y + self.n_a])
z_mean = dgm.forwardPass(self.q_z_axy_mean, l_qz_in)
z_log_var = dgm.forwardPass(self.q_z_axy_log_var, l_qz_in)
z = dgm.sampleNormal(z_mean, z_log_var, mc_samps=n_samples)
z_mean, z_log_var = tf.reshape(
z_mean, [self.mc_samples, -1, self.n_z]), tf.reshape(
z_log_var, [self.mc_samples, -1, self.n_z])
z = tf.reshape(z, [self.mc_samples, -1, self.n_z])
return z_mean, z_log_var, z
def calc_z_prior(self, y):
z_mean_prior = dgm.forwardPass(self.p_z_y_mean, y)
z_log_var_prior = dgm.forwardPass(self.p_z_y_log_var, y)
return z_mean_prior, z_log_var_prior