Action Selection in Bayesian Reinforcement Learning

Abstract

My research attempts to address on-line action selection in reinforcement learning from a Bayesian perspective. The idea is to develop more effective action selection techniques by exploiting information in a Bayesian pos-terior, while also selecting actions by growing an adap-tive, sparse lookahead tree. I further augment the ap-proach by considering a new value function approxima-tion strategy for the belief-state Markov decision pro-cesses induced by Bayesian learning. Bayesian Reinforcement Learning Imagine a mobile vendor robot (“vendorbot”) loaded with snacks and bustling around a building, learning where to visit to optimize its profit. The robot must choose wisely between selling snacks somewhere far away from its home