DMCA
Tunable and Generic Problem Instance Generation for Multi-objective Reinforcement Learning
BibTeX
@MISC{Garrett_tunableand,
author = {Deon Garrett and Jordi Bieger and Kristinn R. Thórisson},
title = {Tunable and Generic Problem Instance Generation for Multi-objective Reinforcement Learning},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract—A significant problem facing researchers in rein-forcement learning, and particularly in multi-objective learning, is the dearth of good benchmarks. In this paper, we present a method and software tool enabling the creation of random problem instances, including multi-objective learning problems, with specific structural properties. This tool, called Merlin (for Multi-objective Environments for Reinforcement LearnINg), pro-vides the ability to control these features in predictable ways, thus allowing researchers to begin to build a more detailed understanding about what features of a problem interact with a given learning algorithm to improve or degrade the algorithm’s performance. We present this method and tool, and briefly discuss the controls provided by the generator, its supported options, and their implications on the generated benchmark instances. I.
Keyphrases
generic problem instance generation multi-objective reinforcement learning rein-forcement learning random problem instance specific structural property multi-objective learning problem learning algorithm benchmark instance reinforcement learning good benchmark algorithm performance predictable way significant problem multi-objective environment detailed understanding multi-objective learning software tool problem interact
