Abstract

In this work we explore a new framework for approximate Bayesian inference inlarge datasets based on stochastic control. We advocate stochastic control as afinite time and low variance alternative to popular steady-state methods suchas stochastic gradient Langevin dynamics (SGLD). Furthermore, we discuss andadapt the existing theoretical guarantees of this framework and establishconnections to already existing VI routines in SDE-based models.

 

Quick Read (beta)