Title :

Inferential Induction: A Novel Framework for Bayesian Reinforcement Learning

Author(s) :

Jorge, Emilio [Auteur]
Chalmers University of Technology [Gothenburg, Sweden]
Eriksson, Hannes [Auteur]
Chalmers University of Technology [Gothenburg, Sweden]
Dimitrakakis, Christos [Auteur]
Chalmers University of Technology [Gothenburg, Sweden]
University of Oslo [UiO]
Basu, Debabrota [Auteur]
Scool [Scool]
Chalmers University of Technology [Gothenburg, Sweden]
Grover, Divya [Auteur]
Chalmers University of Technology [Gothenburg, Sweden]

Conference title :

"I Can't Believe It's Not Better!" at NeurIPS Workshops

City :

Vancouver

Country :

Canada

Start date of the conference :

2020-12-07

Book title :

"I Can't Believe It's Not Better!" at NeurIPS Workshops

Journal title :

Proceedings of Machine Learning Research

Publication date :

2020-12-12

English keyword(s) :

Reinforcement Learning
Bayesian Reinforcement Learning
Policy Evaluation

HAL domain(s) :

Informatique [cs]/Apprentissage [cs.LG]
Informatique [cs]/Intelligence artificielle [cs.AI]
Mathématiques [math]/Statistiques [math.ST]
Informatique [cs]

English abstract : [en]

Bayesian Reinforcement Learning (BRL) offers a decision-theoretic solution to the reinforcement learning problem. While “model-based” BRL algorithms have focused either on maintaining a posterior distribution on models, ...
Show more >Bayesian Reinforcement Learning (BRL) offers a decision-theoretic solution to the reinforcement learning problem. While “model-based” BRL algorithms have focused either on maintaining a posterior distribution on models, BRL “model-free” methods try to estimate value function distributions but make strong implicit assumptions or approximations. We describe a novel Bayesian framework, inferential induction, for correctly inferring value function distributions from data, which leads to a new family of BRL algorithms. We design an algorithm, Bayesian Backwards Induction (BBI), with this framework. We experimentally demonstrate that BBI is competitive with the state of the art. However, its advantage relative to existing BRL model-free methods is not as great as we have expected, particularly when the additional computational burden is taken into account.Show less >

Language :

Anglais

Peer reviewed article :

Oui

Audience :

Internationale

Popular science :

Non

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL

Submission date :

2021-11-13T02:51:44Z