A Meta-Game Evaluation Framework for Deep Multiagent Reinforcement Learning
A Meta-Game Evaluation Framework for Deep Multiagent Reinforcement Learning
About this item
Full title
Author / Creator
Publisher
Ithaca: Cornell University Library, arXiv.org
Journal title
Language
English
Formats
Publication information
Publisher
Ithaca: Cornell University Library, arXiv.org
Subjects
More information
Scope and Contents
Contents
Evaluating deep multiagent reinforcement learning (MARL) algorithms is complicated by stochasticity in training and sensitivity of agent performance to the behavior of other agents. We propose a meta-game evaluation framework for deep MARL, by framing each MARL algorithm as a meta-strategy, and repeatedly sampling normal-form empirical games over c...
Alternative Titles
Full title
A Meta-Game Evaluation Framework for Deep Multiagent Reinforcement Learning
Authors, Artists and Contributors
Author / Creator
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_3049907842
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_3049907842
Other Identifiers
E-ISSN
2331-8422
