TensorFlow 2.0 Beta is available Learn more

Module: tfp.vi

Methods and objectives for variational inference.

Defined in python/vi/__init__.py.

Functions

amari_alpha(...): The Amari-alpha Csiszar-function in log-space.

arithmetic_geometric(...): The Arithmetic-Geometric Csiszar-function in log-space.

chi_square(...): The chi-Square Csiszar-function in log-space.

csiszar_vimco(...): Use VIMCO to lower the variance of gradient[csiszar_function(Avg(logu))].

csiszar_vimco_helper(...): Helper to csiszar_vimco; computes log_avg_u, log_sooavg_u.

dual_csiszar_function(...): Calculates the dual Csiszar-function in log-space.

jeffreys(...): The Jeffreys Csiszar-function in log-space.

jensen_shannon(...): The Jensen-Shannon Csiszar-function in log-space.

kl_forward(...): The forward Kullback-Leibler Csiszar-function in log-space.

kl_reverse(...): The reverse Kullback-Leibler Csiszar-function in log-space.

log1p_abs(...): The log1p-abs Csiszar-function in log-space.

modified_gan(...): The Modified-GAN Csiszar-function in log-space.

monte_carlo_csiszar_f_divergence(...): Monte-Carlo approximation of the Csiszar f-Divergence.

pearson(...): The Pearson Csiszar-function in log-space.

squared_hellinger(...): The Squared-Hellinger Csiszar-function in log-space.

symmetrized_csiszar_function(...): Symmetrizes a Csiszar-function in log-space.

t_power(...): The T-Power Csiszar-function in log-space.

total_variation(...): The Total Variation Csiszar-function in log-space.

triangular(...): The Triangular Csiszar-function in log-space.