Bayesian Policy Search for Stochastic Domains

TL;DR

This paper introduces a Bayesian inference approach for policy search in stochastic domains using nested probabilistic programs and adapts Lightweight Metropolis-Hastings for robust inference, demonstrating effective policy learning.

Contribution

It presents a novel Bayesian formulation for policy search in stochastic domains with nested conditioning and adapts LMH for this setting, broadening inference capabilities.

Findings

Policies of similar quality are learned with the new scheme

The adapted LMH is simpler and more general

The approach is applicable to a wider class of probabilistic programs

Abstract

AI planning can be cast as inference in probabilistic models, and probabilistic programming was shown to be capable of policy search in partially observable domains. Prior work introduces policy search through Markov chain Monte Carlo in deterministic domains, as well as adapts black-box variational inference to stochastic domains, however not in the strictly Bayesian sense. In this work, we cast policy search in stochastic domains as a Bayesian inference problem and provide a scheme for encoding such problems as nested probabilistic programs. We argue that probabilistic programs for policy search in stochastic domains should involve nested conditioning, and provide an adaption of Lightweight Metropolis-Hastings (LMH) for robust inference in such programs. We apply the proposed scheme to stochastic domains and show that policies of similar quality are learned, despite a simpler and more general inference algorithm. We believe that the proposed variant of LMH is novel and applicable to a wider class of probabilistic programs with nested conditioning.