About Me

I am a Postdoctoral Research Associate in the Department of Mathematics at the University of Oxford, and a member of the Erlangen AI Hub. I completed my Ph.D. in Industrial Engineering & Operations Research at UC Berkeley in Summer 2025, advised by Professor Xin Guo. Prior to pursuing Ph.D., I earned a B.S. degree in Mathematics at New York University Shanghai in 2020.

My research interests are at the intersection of stochastic control, reinforcement learning, game theory, and machine learning. More specifically, my research tries to address the following challenges:

• Curse of Many Agents in Stochastic Dynamic Systems:
  Many systems are inherently stochastic and dynamic, and involve multiple agents that interact with each other. To address this, we develop a game-theoretic framework called α-potential games, employing stochastic control approaches and designing multi-agent reinforcement learning (MARL) algorithms for equilibrium analysis and policy learning.

• Lack of Generalizability and Robustness:
  Learned policies and models may fail to generalize to new tasks or withstand adversarial perturbations. Our work studies transfer learning in reinforcement learning (RL) to enable efficient knowledge transfer across related tasks. In parallel, we study machine learning (ML) with adversarial attacks to investigate the robustness and model stability.

• Inefficient Feature Identification in Data Streams:
  Decision-making in high-dimensional systems with limited data remains a major challenge. We address this by exploring rough path methods to extract features from data streams, combining them with ML models for improved prediction and decision-making.

For more detailed research projects, please refer to the Publications tab.

Please find my [CV] here.

Office: S3.20 Andrew Wiles Building, Woodstock Rd, Oxford, UK, OX2 6GG

Email: xinyu.li@maths.ox.ac.uk, xinyu_li@berkeley.edu

My Oxford profile: website