Direct policy search reinforcement learning based on particle filtering
File(s)
Accepted version
Author(s)
Kormushev, Petar
Caldwell, Darwin G
Type
Conference Paper
Abstract
We reveal a link between particle filtering methods and direct policy search reinforcement learning, and propose a novel reinforcement learning algorithm, based heavily on ideas borrowed from particle filters. A major advantage of the proposed algorithm is its ability to perform global search in policy space and thus find the globally optimal policy. We validate the approach on one- and two-dimensional problems with multiple optima, and compare its performance to a global random sampling method, and a state-of-the-art ExpectationMaximization based reinforcement learning algorithm.
Date Issued
2012-06
Date Acceptance
2012-06-30
Citation
The 10th European Workshop on Reinforcement Learning (EWRL 2012), part of the Intl Conf. on Machine Learning (ICML 2012), 2012
Journal / Book Title
The 10th European Workshop on Reinforcement Learning (EWRL 2012), part of the Intl Conf. on Machine Learning (ICML 2012)
Copyright Statement
© 2012 The Authors
Source
The 10th European Workshop on Reinforcement Learning (EWRL 2012)
Publication Status
Published
Start Date
2012-06-30
Finish Date
2012-07-01
Coverage Spatial
Edinburgh, Scotland