Bayesian Reward Filtering - CentraleSupélec Access content directly
Conference Papers Year : 2008

Bayesian Reward Filtering

Olivier Pietquin

Abstract

A wide variety of function approximation schemes have been applied to reinforcement learning. However, Bayesian filtering approaches,which have been shown efficient in other fields such as neural network training, have been little studied.We propose a general Bayesian filtering framework for reinforcement learning, as well as a specific implementation based on sigma point Kalman filtering and kernel machines. This allows us to derive an efficient off-policy model-free approximate temporal differences algorithm which will be demonstrated on two simple benchmarks.

Dates and versions

hal-00351282 , version 1 (08-01-2009)

Identifiers

Cite

Matthieu Geist, Olivier Pietquin. Bayesian Reward Filtering. EWRL 2008, Jun 2008, Lille, France. pp.96-109, ⟨10.1007/978-3-540-89722-4_8⟩. ⟨hal-00351282⟩
32 View
0 Download

Altmetric

Share

Gmail Facebook X LinkedIn More