Configurable Mirror Descent: Towards a Unification of Decision Making

Decision-making problems, categorized as single-agent, e.g., Atari, cooperative multi-agent, e.g., Hanabi, competitive multi-agent, e.g., Hold'em poker, and mixed cooperative and competitive, e.g., football, are ubiquitous in the real world. Various methods are proposed to address the specific...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Pengdeng, Li, Shuxin, Yang, Chang, Wang, Xinrun, Hu, Shuyue, Huang, Xiao, Chan, Hau, An, Bo
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Li, Pengdeng
Li, Shuxin
Yang, Chang
Wang, Xinrun
Hu, Shuyue
Huang, Xiao
Chan, Hau
An, Bo
description Decision-making problems, categorized as single-agent, e.g., Atari, cooperative multi-agent, e.g., Hanabi, competitive multi-agent, e.g., Hold'em poker, and mixed cooperative and competitive, e.g., football, are ubiquitous in the real world. Various methods are proposed to address the specific decision-making problems. Despite the successes in specific categories, these methods typically evolve independently and cannot generalize to other categories. Therefore, a fundamental question for decision-making is: \emph{Can we develop \textbf{a single algorithm} to tackle \textbf{ALL} categories of decision-making problems?} There are several main challenges to address this question: i) different decision-making categories involve different numbers of agents and different relationships between agents, ii) different categories have different solution concepts and evaluation measures, and iii) there lacks a comprehensive benchmark covering all the categories. This work presents a preliminary attempt to address the question with three main contributions. i) We propose the generalized mirror descent (GMD), a generalization of MD variants, which considers multiple historical policies and works with a broader class of Bregman divergences. ii) We propose the configurable mirror descent (CMD) where a meta-controller is introduced to dynamically adjust the hyper-parameters in GMD conditional on the evaluation measures. iii) We construct the \textsc{GameBench} with 15 academic-friendly games across different decision-making categories. Extensive experiments demonstrate that CMD achieves empirically competitive or better outcomes compared to baselines while providing the capability of exploring diverse dimensions of decision making.
doi_str_mv 10.48550/arxiv.2405.11746
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2405_11746</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2405_11746</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-7c435c6052c1d2cd9f31b203e70d087ddddeb3a190ce453d8ac124b8eb3d8e673</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwAr_QILfdtmhUB5SKzZhHd1c25VFsZFTXn9P2jKbGY1GIx1CrjhrldOa3UD9SV-tUEy3nFtlzsmqKzmm7WeFcRfoJtVaKr0PE4a8v6V9-YbqJwr0NaeYEPapZFrivMA0HfIG3lLeXpCzCLspXP77gvQPq757atYvj8_d3boBY01jUUmNhmmB3Av0yyj5KJgMlnnmrJ8VRgl8yTAoLb0D5EKNbi69C8bKBbk-3R45ho-a3qH-Dgee4cgj_wCHG0Wu</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Configurable Mirror Descent: Towards a Unification of Decision Making</title><source>arXiv.org</source><creator>Li, Pengdeng ; Li, Shuxin ; Yang, Chang ; Wang, Xinrun ; Hu, Shuyue ; Huang, Xiao ; Chan, Hau ; An, Bo</creator><creatorcontrib>Li, Pengdeng ; Li, Shuxin ; Yang, Chang ; Wang, Xinrun ; Hu, Shuyue ; Huang, Xiao ; Chan, Hau ; An, Bo</creatorcontrib><description>Decision-making problems, categorized as single-agent, e.g., Atari, cooperative multi-agent, e.g., Hanabi, competitive multi-agent, e.g., Hold'em poker, and mixed cooperative and competitive, e.g., football, are ubiquitous in the real world. Various methods are proposed to address the specific decision-making problems. Despite the successes in specific categories, these methods typically evolve independently and cannot generalize to other categories. Therefore, a fundamental question for decision-making is: \emph{Can we develop \textbf{a single algorithm} to tackle \textbf{ALL} categories of decision-making problems?} There are several main challenges to address this question: i) different decision-making categories involve different numbers of agents and different relationships between agents, ii) different categories have different solution concepts and evaluation measures, and iii) there lacks a comprehensive benchmark covering all the categories. This work presents a preliminary attempt to address the question with three main contributions. i) We propose the generalized mirror descent (GMD), a generalization of MD variants, which considers multiple historical policies and works with a broader class of Bregman divergences. ii) We propose the configurable mirror descent (CMD) where a meta-controller is introduced to dynamically adjust the hyper-parameters in GMD conditional on the evaluation measures. iii) We construct the \textsc{GameBench} with 15 academic-friendly games across different decision-making categories. Extensive experiments demonstrate that CMD achieves empirically competitive or better outcomes compared to baselines while providing the capability of exploring diverse dimensions of decision making.</description><identifier>DOI: 10.48550/arxiv.2405.11746</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computer Science and Game Theory ; Computer Science - Learning ; Computer Science - Multiagent Systems</subject><creationdate>2024-05</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2405.11746$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2405.11746$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Li, Pengdeng</creatorcontrib><creatorcontrib>Li, Shuxin</creatorcontrib><creatorcontrib>Yang, Chang</creatorcontrib><creatorcontrib>Wang, Xinrun</creatorcontrib><creatorcontrib>Hu, Shuyue</creatorcontrib><creatorcontrib>Huang, Xiao</creatorcontrib><creatorcontrib>Chan, Hau</creatorcontrib><creatorcontrib>An, Bo</creatorcontrib><title>Configurable Mirror Descent: Towards a Unification of Decision Making</title><description>Decision-making problems, categorized as single-agent, e.g., Atari, cooperative multi-agent, e.g., Hanabi, competitive multi-agent, e.g., Hold'em poker, and mixed cooperative and competitive, e.g., football, are ubiquitous in the real world. Various methods are proposed to address the specific decision-making problems. Despite the successes in specific categories, these methods typically evolve independently and cannot generalize to other categories. Therefore, a fundamental question for decision-making is: \emph{Can we develop \textbf{a single algorithm} to tackle \textbf{ALL} categories of decision-making problems?} There are several main challenges to address this question: i) different decision-making categories involve different numbers of agents and different relationships between agents, ii) different categories have different solution concepts and evaluation measures, and iii) there lacks a comprehensive benchmark covering all the categories. This work presents a preliminary attempt to address the question with three main contributions. i) We propose the generalized mirror descent (GMD), a generalization of MD variants, which considers multiple historical policies and works with a broader class of Bregman divergences. ii) We propose the configurable mirror descent (CMD) where a meta-controller is introduced to dynamically adjust the hyper-parameters in GMD conditional on the evaluation measures. iii) We construct the \textsc{GameBench} with 15 academic-friendly games across different decision-making categories. Extensive experiments demonstrate that CMD achieves empirically competitive or better outcomes compared to baselines while providing the capability of exploring diverse dimensions of decision making.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computer Science and Game Theory</subject><subject>Computer Science - Learning</subject><subject>Computer Science - Multiagent Systems</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwAr_QILfdtmhUB5SKzZhHd1c25VFsZFTXn9P2jKbGY1GIx1CrjhrldOa3UD9SV-tUEy3nFtlzsmqKzmm7WeFcRfoJtVaKr0PE4a8v6V9-YbqJwr0NaeYEPapZFrivMA0HfIG3lLeXpCzCLspXP77gvQPq757atYvj8_d3boBY01jUUmNhmmB3Av0yyj5KJgMlnnmrJ8VRgl8yTAoLb0D5EKNbi69C8bKBbk-3R45ho-a3qH-Dgee4cgj_wCHG0Wu</recordid><startdate>20240519</startdate><enddate>20240519</enddate><creator>Li, Pengdeng</creator><creator>Li, Shuxin</creator><creator>Yang, Chang</creator><creator>Wang, Xinrun</creator><creator>Hu, Shuyue</creator><creator>Huang, Xiao</creator><creator>Chan, Hau</creator><creator>An, Bo</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240519</creationdate><title>Configurable Mirror Descent: Towards a Unification of Decision Making</title><author>Li, Pengdeng ; Li, Shuxin ; Yang, Chang ; Wang, Xinrun ; Hu, Shuyue ; Huang, Xiao ; Chan, Hau ; An, Bo</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-7c435c6052c1d2cd9f31b203e70d087ddddeb3a190ce453d8ac124b8eb3d8e673</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computer Science and Game Theory</topic><topic>Computer Science - Learning</topic><topic>Computer Science - Multiagent Systems</topic><toplevel>online_resources</toplevel><creatorcontrib>Li, Pengdeng</creatorcontrib><creatorcontrib>Li, Shuxin</creatorcontrib><creatorcontrib>Yang, Chang</creatorcontrib><creatorcontrib>Wang, Xinrun</creatorcontrib><creatorcontrib>Hu, Shuyue</creatorcontrib><creatorcontrib>Huang, Xiao</creatorcontrib><creatorcontrib>Chan, Hau</creatorcontrib><creatorcontrib>An, Bo</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Li, Pengdeng</au><au>Li, Shuxin</au><au>Yang, Chang</au><au>Wang, Xinrun</au><au>Hu, Shuyue</au><au>Huang, Xiao</au><au>Chan, Hau</au><au>An, Bo</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Configurable Mirror Descent: Towards a Unification of Decision Making</atitle><date>2024-05-19</date><risdate>2024</risdate><abstract>Decision-making problems, categorized as single-agent, e.g., Atari, cooperative multi-agent, e.g., Hanabi, competitive multi-agent, e.g., Hold'em poker, and mixed cooperative and competitive, e.g., football, are ubiquitous in the real world. Various methods are proposed to address the specific decision-making problems. Despite the successes in specific categories, these methods typically evolve independently and cannot generalize to other categories. Therefore, a fundamental question for decision-making is: \emph{Can we develop \textbf{a single algorithm} to tackle \textbf{ALL} categories of decision-making problems?} There are several main challenges to address this question: i) different decision-making categories involve different numbers of agents and different relationships between agents, ii) different categories have different solution concepts and evaluation measures, and iii) there lacks a comprehensive benchmark covering all the categories. This work presents a preliminary attempt to address the question with three main contributions. i) We propose the generalized mirror descent (GMD), a generalization of MD variants, which considers multiple historical policies and works with a broader class of Bregman divergences. ii) We propose the configurable mirror descent (CMD) where a meta-controller is introduced to dynamically adjust the hyper-parameters in GMD conditional on the evaluation measures. iii) We construct the \textsc{GameBench} with 15 academic-friendly games across different decision-making categories. Extensive experiments demonstrate that CMD achieves empirically competitive or better outcomes compared to baselines while providing the capability of exploring diverse dimensions of decision making.</abstract><doi>10.48550/arxiv.2405.11746</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2405.11746
ispartof
issn
language eng
recordid cdi_arxiv_primary_2405_11746
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computer Science and Game Theory
Computer Science - Learning
Computer Science - Multiagent Systems
title Configurable Mirror Descent: Towards a Unification of Decision Making
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-07T10%3A48%3A00IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Configurable%20Mirror%20Descent:%20Towards%20a%20Unification%20of%20Decision%20Making&rft.au=Li,%20Pengdeng&rft.date=2024-05-19&rft_id=info:doi/10.48550/arxiv.2405.11746&rft_dat=%3Carxiv_GOX%3E2405_11746%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true