Bayesian Neural Networks in tensorflow probability: quick start.

Preamble: Bayesian Neural Networks, allow us to exploit uncertainty and therefore allow us to develop robust models. Thus knowledge of uncertainty is fundamental to development of robust and safe machine learning techniques. This in post we outline the two main types of uncertainties and how to model them using tensorflow probability via simple models. We employ Bayesian framework, which is applicable to deep learning and reinforcement learning. For completeness lets restate baye’s rule: posterior probability is prior probability time the…