Chengshuai Shi


Third-year Ph.D. student
Department of Electrical and Computer Engineering
University of Virginia

Email: cs7ync at virginia dot edu

Phone: 434-218-9860

Advisor: Prof. Cong Shen

About this picture: My mother drew this Dr. Cat wearing a graduation cap when I was in high school. Now I am working towards a Ph.D. and hopefully, my family can see me wearing it one day:)

Research Interest

My current research focuses on multi-armed bandits (MAB) problems and more general reinforcement learning problems. I also have broad interests in distributed learning and federated learning.

Some of my ongoing research projects are

  • Decentralized multi-player MAB (MP-MAB): The MP-MAB problem, in which multiple players simultaneously play the bandit game in a fully decentralized fashion and interact through arm collisions, has sparked significant interest in recent years. I have introduced information-theoretical tools, e.g., error correction/detection coding and Z-channel model, into the study of no-sensing MP-MAB, where collisions are not perceivable by players (stochastic, adversarial). Currently, I am working on opening the methodological door between MP-MAB and combinatorial MAB.

  • Federated MAB (FMAB): Federated learning (FL) is an emerging and promising distributed machine learning paradigm. While the current FL studies are mostly focused on supervised learning, I proposed to extend core principles in FL into the study of MAB, and am actively working on this new bandit paradigm (paper, paper).


  • 09/2021: Two papers accepted to NeurIPS 2021!

    • The first one is on the heterogeneous MP-MAB problem, titled “Heterogeneous Multi-player Multi-armed Bandits: Closing the Gap and Generalization”;

    • The other one is about best arm identification in FMAB, titled “(Almost) Free Incentivized Exploration from Decentralized Learning Agents”.

  • 02/2021: Paper accepted as an oral presentation at AISTATS 2021!

    • Our paper titled “Federated Multi-armed Bandits with Personalization” (with Prof. Cong Shen and Prof. Jing Yang) is accepted as an oral presentation at AISTATS 2021;

    • This is quite an honor for me, especially in such a competitive year (approx 3%, 48 oral | 455 accepted | 1527 submitted).