Junyu Xuan

About

Junyu Xuan is a rising researcher at the forefront of reinforcement learning (RL), specializing in making intelligent agents robust and adaptable in complex, real-world settings. His work tackles two critical challenges: enabling RL to function in nonstationary environments where conditions shift unpredictably, and improving the sample efficiency of deep reinforcement learning (DRL). In his highly cited 2018 paper, "Bayesian Deep Reinforcement Learning via Deep Kernel Learning" (13 citations), Xuan introduced a novel framework that merges Bayesian inference with deep kernel methods, allowing agents to quantify uncertainty and learn more effectively from limited data. Building on this, his 2024 work, "Deep Reinforcement Learning in Nonstationary Environments With Unknown Change Points" (8 citations), directly addresses the practical hurdle of time-varying state transitions, proposing a method to detect and adapt to changes without prior knowledge—a breakthrough for applications like robotics and autonomous systems. His 2023 paper on "Transformed Successor Features" further advances transfer learning, enabling agents to reuse knowledge across tasks. Xuan’s contributions are pivotal for deploying RL in dynamic, unpredictable domains, marking him as a key innovator in bridging theory and real-world deployment.

Research Focus

Key Achievements

Top Papers

1. 1
   Bayesian Deep Reinforcement Learning via Deep Kernel Learning

   13 citations · 2018
2. 2
   Deep Reinforcement Learning in Nonstationary Environments With Unknown Change Points

   8 citations · 2024
3. 3
   Transformed Successor Features for Transfer Reinforcement Learning

   3 citations · 2023

Key Collaborators

Contact & Links