The aim of the Multi-Agent Reinforcement Learning (MARL) reading group at the Computer Science Department of the University of Maryland, College Park is to pursue foundational research discussions, collaborations and brainstorming on MARL research frameworks with application areas like self-driving cars, solving game theory problems like discovering, analyzing and disrupting illicit trafficking networks (be it arms, humans, wildlife, rare plants and deforested trees, illegal drugs, counterfeit goods), war games and strategies, equitable markets, fair tax policies, swarm robotics, climate change and existential disasters like pandemics among other domains. We focus on different modalities of Multi-agent Reinforcement Learning like cooperation and competition which are inspired from game theory, online learning, meta learning, curriculum learning among other Machine Learning areas which can be used to solve real world problems. We consider the fundamental Human-AI alignment problem through Multi-Agent Reinforcement Learning to be pertinent for discussions in our reading group.
Our group is collaborative and diverse comprising of participants from academia and industry like the University of Maryland, College Park (UMD); Carnegie Mellon University (CMU); University of California, Berkeley (UCB); Massachussetts Institute of Technology (MIT); Harvard University; University of Illinois, Urbana Champaign (UIUC); Google DeepMind; Google Brain; Google X; Google Research; Columbia University; Virginia Tech; University of Virginia; Facebook AI Research (FAIR); Microsoft Research; Micron Technology; Amazon; Boeing; Stanford University; University of Chicago; New York University (NYU); University of California, San Diego (UCSD); Boston University; University of Minnesota (UMN); University of Oxford; Technical University of Munich (TU Munich); Delft University of Technology (TU Delft); Autonomous University of Madrid (UAM); Institute of Mathematical Sciences (ICMAT) Madrid; Indian Institute of Technology (IIT) Kanpur; Indian Institute of Technology (IIT) Kharagpur; and Korea Advanced Institute of Science and Technology (KAIST).
The reading group usually takes place at 4pm ET/1 pm PT on Thursdays. Timings are updated every semester based on a when2meet, to consider participant inputs.
Participants can join on Google Meet or in-person from IRB-5105 at the Brendan Iribe Center for Computer Science and Engineering, University of Maryland, College Park.
We have an initial list of papers in the domain in the Google Drive. All slides are shared and reading group presentations are recorded and shared on a Google Sheets in the Google Drive in case anyone cannot join to facilitate asynchronous discussion on Google Groups and Slack. Talks are normally 45 minutes to 1 hour focusing on the presentation, Q&A and brainstorming to apply ideas relevant to current research.
If you're interested in joining the group, please fill out the Google Form below. We will invite you to our Google group and Slack group. For queries, further information, or to request an invitation to join the group or to give a talk, please email Saptarashmi Bandyopadhyay (saptab1@umd.edu).