f_mean = tf.matmul(K, self.q_alpha) + self.mean_function(self.X)
// for each of the data-dimensions (columns of Y), find the diagonal
// of the variance, and also relevant parts of the KL.
f_var = [], tf.zeros((1,), tf.float64), tf.zeros((1,), tf.float64)
A_logdet = tf.zeros((1,), tf.float64)
trAi = tf.zeros((1,), tf.float64)
for d in range(self.num_latent):
b = self.q_lambda[:, d]
After Change
f_mean = tf.matmul(K, self.q_alpha) + self.mean_function(self.X)
// for each of the data-dimensions (columns of Y), find the diagonal
// of the variance, and also relevant parts of the KL.
f_var = []
A_logdet = tf.zeros((1,), tf.float64)
trAi = tf.zeros((1,), tf.float64)
for d in range(self.num_latent):
b = self.q_lambda[:, d]
