Skip to Main content Skip to Navigation
Conference papers

Is the Bellman residual a bad proxy?

Matthieu Geist 1 Bilal Piot 2 Olivier Pietquin 3
3 SEQUEL - Sequential Learning
Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
Abstract : This paper aims at theoretically and empirically comparing two standard optimization criteria for Reinforcement Learning: i) maximization of the mean value and ii) minimization of the Bellman residual. For that purpose, we place ourselves in the framework of policy search algorithms, that are usually designed to maximize the mean value, and derive a method that minimizes the residual T * v π − v π 1,ν over policies. A theoretical analysis shows how good this proxy is to policy optimization , and notably that it is better than its value-based counterpart. We also propose experiments on randomly generated generic Markov decision processes, specifically designed for studying the influence of the involved concentrability coefficient. They show that the Bellman residual is generally a bad proxy to policy optimization and that directly maximizing the mean value is much better, despite the current lack of deep theoretical analysis. This might seem obvious, as directly addressing the problem of interest is usually better, but given the prevalence of (projected) Bellman residual minimization in value-based reinforcement learning, we believe that this question is worth to be considered.
Document type :
Conference papers
Complete list of metadata

Cited literature [23 references]  Display  Hide  Download
Contributor : Matthieu Geist Connect in order to contact the contributor
Submitted on : Monday, November 6, 2017 - 5:15:32 PM
Last modification on : Wednesday, October 27, 2021 - 7:06:21 AM


Files produced by the author(s)


  • HAL Id : hal-01629739, version 1


Matthieu Geist, Bilal Piot, Olivier Pietquin. Is the Bellman residual a bad proxy?. NIPS 2017 - Advances in Neural Information Processing Systems, Dec 2017, Long Beach, United States. pp.1-13. ⟨hal-01629739⟩



Record views


Files downloads