OptLearnMAS-23
The 14th Workshop on Optimization and Learning in Multiagent Systems
at AAMAS 2023 at London, United Kingdom
Stimulated by various emerging applications involving agents to solve complex problems in
real-world domains, such as intelligent sensing systems for the Internet of the Things (IoT),
automated configurators for critical infrastructure networks, and intelligent resource allocation for
social domains (e.g., security games for the deployment of security resources or
auctions/procurements for allocating goods and services), agents in these domains commonly
leverage different forms optimization and/or learning to solve complex problems.
The goal of the workshop is to provide researchers with a venue to discuss models or techniques for
tackling a variety of multi-agent optimization problems. We seek contributions in the general area of
multi-agent optimization, including distributed optimization, coalition formation, optimization
under uncertainty, winner determination algorithms in auctions and procurements, and algorithms
to compute Nash and other equilibria in games. Of particular emphasis are contributions at the
intersection of optimization and learning.
See below for a (non-exhaustive) list of topics.
This workshop invites works from different strands of the multi-agent systems community that pertain to the design of algorithms, models, and techniques to deal with multi-agent optimization and learning problems or problems that can be effectively solved by adopting a multi-agent framework.
The workshop is of interest both to researchers investigating applications of multi-agent systems to optimization problems in large, complex domains, as well as to those examining optimization and learning problems that arise in systems comprised of many autonomous agents. In so doing, this workshop aims to provide a forum for researchers to discuss common issues that arise in solving optimization and learning problems in different areas, to introduce new application domains for multi-agent optimization techniques, and to elaborate common benchmarks to test solutions.
Finally, the workshop will welcome papers that describe the release of benchmarks and data sets that can be used by the community to solve fundamental problems of interest, including in machine learning and optimization for health systems and urban networks, to mention but a few examples.
The workshop will be a one-day meeting. It will include a number of technical sessions, a poster session where presenters can discuss their work, with the aim of further fostering collaborations, multiple invited speakers covering crucial challenges for the field of multiagent optimization and learning.
Time (UK) GMT+1 | Talk | |
---|---|---|
8:30 | Introductory remarks | |
8:40 | Invited Talk 1 by Erich Wehrle: "Challenges in Design Space Exploration and Design Optimization of Aerospace Systems" | |
Session: Optimization and Learning -- Session chair: Jiaoyang Li | ||
9:30 | Contributed Talk: Asynchronous Communication Aware Multi-Agent Task Allocation | |
9:45 | Contributed Talk: A Multiagent Path Search Algorithm for Large-Scale Coalition Structure Generation | |
10:00 | Coffee Break | |
Session 2: Optimization and Learning -- Session chair: Hau Chan | ||
10:45 | Contributed Talk: A DCOP Approach to Distributed Multi-Agent Pathfinding | |
11:00 | Contributed Talk: A Novel Degree Planning Approach based on Mixed-Integer Programming | |
11:15 | Contributed Talk: Faster Optimal Coalition Structure Generation via Offline Coalition Selection and Graph-Based Search | |
11:30 | Contributed Talk: A Norm Approximation Approach for Research Project Planning | |
11:45 | Contributed Talk: Towards a Unifying Model of Rationality in Multiagent Systems | |
12:00 | Contributed Talk: Cooperative Personalized Bilevel Optimization over Random Directed Networks | |
12:15 | Contributed Talk: Coordinating Fully-Cooperative Agents Using Hierarchical Learning Anticipation | |
12:30 | Lunch Break | |
14:00 | Invited Talk by Georgios Piliouras: "Towards a Unified Theory of Learning in Games: Chaos, Anarchy, Regret & Equilibration" | |
Session 3: Optimization and Learning with Strategic Agents -- Session chair: James Kotary | ||
14:45 | Contributed Talk: Efficient Size-based Hybrid Algorithm for Optimal Coalition Structure Generation | |
15:00 | Contributed Talk: GENEPI: a GENEric Parameter-sharing for Intrinsically motivated MARL agents | |
15:15 | Contributed Talk: Bayesian Rationality in Satisfaction Games | |
15:30 | Contributed Talk: The Computational Complexity of Single-Player Imperfect-Recall Games | |
15:45 | Coffee Break | |
Session 4: Optimization and Learning with Strategic Agents -- Session chair: Filippo Bistaffa | ||
16:30 | Contributed Talk: Learning not to Regret | |
16:45 | Contributed Talk: Pareto Actor-Critic for Equilibrium Selection in Multi-Agent Reinforcement Learning | |
17:00 | Contributed Talk: Ascending-Price Mechanism for General Multi-Sided Markets | |
17:15 | Contributed Talk: A Recommendation System for Participatory Budgeting | |
17:30 | Contributed Talk: Promoting Non-Cooperation Through Ordering | |
17:45 | Closing Remarks |
Challenges in design space exploration and design optimization of aerospace systems
Abstract:
New requirements for sustainable, efficient and economic aviation present the sector with new design possibilities and constraints.
The aviation industry has committed to net-zero carbon emissions by 2050.
This ambitious goal necessitates a new thinking in aircraft design to accelerate development with special attention to these goals in the development at aircraft as well as system and component levels.
Algorithmic-supported design methods in concert with accurate parametric models are powerful tools in the conception and development of aerospace systems and components.
As such, applications range from optimal design of aircraft system architecture and high-performance structures of fiber-reinforced polymer to optimal parameters for manufacturing systems to optimal flight path design.
This talk will discuss reason-based system architecture exploration and design optimization of complex aircraft systems.
As central role of multidisciplinary design optimization, this includes the use of physics-based parametric modeling and advanced multi-fidelity surrogating.
The challenges to application of optimization and reason-based design exploration will be outlined and needs for future research in this area motivated.
Bio:
Dr.-Ing. Erich Wehrle leads the Multidisciplinary Design Optimization research group at Collins Aerospace.
The team’s research focuses on algorithmic support in the design of complex aircraft systems, especially via system architecture exploration and design optimization considering multiple disciplines and physics-based simulation models.
He previously held roles in academia as Assistant Professor at the Free University of Bozen-Bolzano (Italy) and as postdoctoral researcher at the Technical University Munich (Germany) in which his research addressed design optimization of lightweight structures and mechanical systems under a multidisciplinary perspective, including structural analysis, multibody dynamics, crash mechanics and uncertainty.
He holds a doctorate and a master in Mechanical Engineering from the Technical University of Munich and a bachelor in Mechanical Engineering from the State University of New York at Buffalo (USA).
Towards a Unified Theory of Learning in Games: Chaos, Anarchy, Regret & Equilibration
Abstract:
We examine some classic questions in game theory and online learning. How do standard online optimization, regret-minimizing dynamics such as multiplicative weights update, online gradient descent, and follow-the-regularized-leader (and variants thereof) behave when applied in games?
The standard textbook answer to this question is that these dynamics converge in a time-average sense to weak notions of equilibria. We discuss weaknesses of such results in practice and turn our focus towards understanding the day-to-day behavior instead.
In the seminal class of zero-sum games, continuous-time dynamics ``cycle" around the equilibrium, whereas discrete time-algorithms can behave in all possible ways (equilibration, recurrence/cycles, or unstable chaos) based only on minor changes.
In the standard class of potential/congestion games, such dynamics experience phase transitions from stability to chaos as the total system demand increases. Such instabilities lead to high social inefficiency even in games where the Price of Anarchy is equal to one (i.e. where all equilibria are optimal).
Our analysis is derived by a combination of distinct tools: optimization theory (regret analysis), dynamical systems (chaos theory) and game theory (equilibria, Price of Anarchy).
Bio:
Georgios Piliouras is a senior research scientist at Google DeepMind and an associate professor at the Singapore University of Technology and Design (on leave). His research interests lie in the areas of multi-agent learning, algorithmic game theory, blockchain and dynamical systems.
He received his PhD in Computer Science from Cornell University.
He has held research positions at the Georgia Institute of Technology (GaTech, ECE Dept.) and California Institute of Technology (Caltech, Dept. of Computing and Mathematical Sciences) as well as a visiting professor position at UC Berkeley.
He is the recipient of a Singapore NRF Fellowship (2018) and a Simons/UC Berkeley Fellowship (2015).
His work has been recognized with several distinctions including the outstanding paper award at AAAI (2021) and best paper runner-up at AAMAS (2019, 2022).
Submission URL: https://cmt3.research.microsoft.com/OptLearnMAS2023/Submission/Index
All papers must be submitted in PDF format, using the AAMAS-23 author kit.
Submissions should include the name(s), affiliations, and email addresses of all authors.
Submissions will be refereed on the basis of technical quality, novelty, significance, and
clarity. Each submission will be thoroughly reviewed by at least two program committee members.
Submissions of papers rejected from the IJCAI 2023 technical program are welcomed.
Rejected IJCAI papers with *average* scores of at least weak reject/weak accept may be submitted
to OptLearnMAS along with previous reviews and scores and an optional letter indicating how the
authors have addressed the reviewers comments.
Please use the submission link above and indicate that the submission is a resubmission from
of a IJCAI rejected paper. Also OptLearnMAS submission, reviews and optional letter
need to be compiled into a single pdf file.
These submissions will not undergo the regular review process, but a light one, performed by the
chairs, and will be accepted if the previous reviews are
judged to meet the workshop standard.
Per the AAMAS Workshop organizers:
There will be a Springer issue for best workshop papers and visionary papers, so each workshop should nominate two papers, one for each special issue. The authors should be aware that if the nominated workshop paper is also an AAMAS paper (or some other conference paper), the version in the Springer books should have additional material (at least 30% more).
For questions about the submission process, contact the workshop chairs.