Variational Inference

Bayesian methods require assumptions that are often misaligned with the realities of modern day inference. It is my conviction that we can overcome these challenges with an optimisation-centric generalisation of Bayes' rule.