Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

University of Alberta

$20,000.00

Learning and Search in Decision Domains Featuring Large Action Sets and Uncertainty

EF 2017-18

Résumé
Artificial Intelligence (AI) research has come a long way creating systems that challenge human supremacy in complex decision domains such as Chess, Poker, Jeopardy, stock trading, and recently image recognition, autonomous driving, and the Asian boardgame Go. By contrast, AI progress in popular video games (such as StarCraft) and complex imperfect information domains, in which information is hidden from agents, has been slow, and in many cases human experts can still easily outperform the best programs. The human advantage in these domains can in part be attributed to our abilities to simplify problems while maintaining solutions, to search at different abstraction levels (e.g., looking into details only when high-level solution concepts don't seem to work), to adjust behaviour to opponents and partners, and to infer intentions from observed actions. Machines achieving human proficiency in such areas will have a huge impact on technology, ranging from creating effective multi-robot coordination in construction tasks and rescue missions to automating naval battle groups, which can greatly benefit the Canadian economy and security. In this proposal I will lay out a research plan aiming at closing human-machine performance gaps in decision domains with large structured action sets, in which classical search algorithms fail due to combinatorial explosion (Theme 1), and in domains featuring agent cooperation and uncertainty created by hiding information from agents (Theme 2). The focus of the proposed work is to explore how to better integrate heuristic search (which can evaluate the merit of actions by looking ahead) with machine learning to deal with large combinatorial action spaces, imperfect information, and agent cooperation. In these areas the gained general insights will be used to develop AI systems able to compete with human experts in challenging decision domains, which include modern video games, military combat simulations, and traffic light optimization for Theme 1, and team games, such as Contract Bridge, and combat support systems with human operators in the loop for Theme 2. There are numerous application areas that can benefit from the proposed work. Video game companies, for instance, can create AI systems that improve gaming experiences by adjusting to players based on analyzing game records, or byusing stronger AI systems to fine-tune parameters via self-play. Likewise, learning to orchestrate dozens of acting entities in real-time settings can improve traffic flow, military training simulators, as well as distributedweapon systems, and being able to learn to cooperate with humans will also be important in the future, with robots becoming ever more autonomous.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
RGPIN-2017-05271
Name nom
Buro, Michael
Department département
Computing Science - Computing Science
Organization id
9
Institution établissement
University of Alberta
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2017
Competition year année de concours
2017
Award amount
20000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2800
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Artificial intelligence (computer vision, use2603)
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Partie
1
Nb partie
1
Application title
Learning and Search in Decision Domains Featuring Large Action Sets and Uncertainty
Keyword
NSERC
Application summary
Artificial Intelligence (AI) research has come a long way creating systems that challenge human supremacy in complex decision domains such as Chess, Poker, Jeopardy, stock trading, and recently image recognition, autonomous driving, and the Asian boardgame Go. By contrast, AI progress in popular video games (such as StarCraft) and complex imperfect information domains, in which information is hidden from agents, has been slow, and in many cases human experts can still easily outperform the best programs. The human advantage in these domains can in part be attributed to our abilities to simplify problems while maintaining solutions, to search at different abstraction levels (e.g., looking into details only when high-level solution concepts don't seem to work), to adjust behaviour to opponents and partners, and to infer intentions from observed actions. Machines achieving human proficiency in such areas will have a huge impact on technology, ranging from creating effective multi-robot coordination in construction tasks and rescue missions to automating naval battle groups, which can greatly benefit the Canadian economy and security. In this proposal I will lay out a research plan aiming at closing human-machine performance gaps in decision domains with large structured action sets, in which classical search algorithms fail due to combinatorial explosion (Theme 1), and in domains featuring agent cooperation and uncertainty created by hiding information from agents (Theme 2). The focus of the proposed work is to explore how to better integrate heuristic search (which can evaluate the merit of actions by looking ahead) with machine learning to deal with large combinatorial action spaces, imperfect information, and agent cooperation. In these areas the gained general insights will be used to develop AI systems able to compete with human experts in challenging decision domains, which include modern video games, military combat simulations, and traffic light optimization for Theme 1, and team games, such as Contract Bridge, and combat support systems with human operators in the loop for Theme 2. There are numerous application areas that can benefit from the proposed work. Video game companies, for instance, can create AI systems that improve gaming experiences by adjusting to players based on analyzing game records, or byusing stronger AI systems to fine-tune parameters via self-play. Likewise, learning to orchestrate dozens of acting entities in real-time settings can improve traffic flow, military training simulators, as well as distributedweapon systems, and being able to learn to cooperate with humans will also be important in the future, with robots becoming ever more autonomous.