Multi-agent Reinforcement Learning: Theory and Applications
Interdisciplinary Areas: | Data/Information/Computation |
---|
Project Description
Reinforcement learning has been increasingly adopted for decentralized, autonomous decision making in many fields, ranging from the operation of cyber-physical systems (e.g., transportation control) and telecommunication networks (e.g., opportunistic spectrum access) to cooperative robots in the production factories and social networks. Mounting recent empirical evidence suggests that there exists substantial risk of unfairness in the learning algorithms and models. As real-world problems in these applications are becoming more complex, there are many situations that a single learning agent or a monolithic system is not able to cope with, mandating the use of multi-agent learning to yield the best results. Unfortunately, existing notions/tools of fairness do not generalize well to these settings, where multiple intelligent agents—acting competitively, cooperatively, or neutrally—dynamically interact with an environment that is subjected to the actions of all agents. One key challenge in multi-agent framework is that the collaborative decisions optimize a function of long term reward of the agents which even for centralized decision making cannot be performed with single agent frameworks.
Start Date
May 2020
Postdoc Qualifications
The postdoc must have experience in theoretical aspects of data science, with a preferred PhD in CS/ECE/Stats/Math or related areas.
Co-advisors
Vaneet Aggarwal
vaneet@purdue.edu
School of IE and ECE (by courtesy)
https://engineering.purdue.edu/CLANLabs
Chris Brinton
cgb@purdue.edu
School of ECE
http://www.cbrinton.net/
References
Buşoniu, Lucian, Robert Babuška, and Bart De Schutter. "Multi-agent reinforcement learning: An overview." Innovations in multi-agent systems and applications-1. Springer, Berlin, Heidelberg, 2010. 183-221.
Aggarwal et al., Reinforcement Learning for Mean Field Game, arXiv preprint arXiv:1905.13357
Aggarwal et al., A systematic framework of fairness with multiple agents, ICASSP 2020.