Tuning local search by average-reward reinforcement learning

Research output: Chapter in Book/Report/Conference proceedingsChapterpeer-review

Abstract

Reinforcement Learning and local search have been combined in a variety of ways, in order to learn how to solve combinatorial problems more efficiently. Most approaches optimise the total reward, where the reward at each action is the change in objective function. We argue that it is more appropriate to optimise the average reward. We use R-learning to dynamically tune noise in standard SAT local search algorithms on single instances. Experiments show that noise can be successfully automated in this way.

Original languageEnglish
Title of host publicationLearning and Intelligent Optimization - Second International Conference, LION 2007 II, Selected Papers
Pages192-205
Number of pages14
DOIs
Publication statusPublished - 2008
Event2nd International Conference on Learning and Intelligent Optimization, LION 2007 II - Trento, Italy
Duration: 8 Dec 200712 Dec 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5313 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2nd International Conference on Learning and Intelligent Optimization, LION 2007 II
Country/TerritoryItaly
CityTrento
Period8/12/0712/12/07

Fingerprint

Dive into the research topics of 'Tuning local search by average-reward reinforcement learning'. Together they form a unique fingerprint.

Cite this