Henghui Zhu

About

Henghui Zhu is a researcher specializing in reinforcement learning and data-driven decision-making under uncertainty, with a focus on Markov decision processes (MDPs). His work bridges theoretical foundations and practical algorithms for learning optimal policies directly from observed data, without requiring explicit knowledge of system dynamics. In his highly cited 2018 paper, Zhu introduced a framework for learning policies in MDPs by parameterizing them using state-action features—either handcrafted or kernel-based—enabling efficient policy extraction from trajectory data. This contribution has been foundational for researchers working on imitation learning and offline reinforcement learning. His 2020 work extended this approach by jointly learning parametric policies and transition probability models, further advancing the ability to model and control stochastic systems from limited data. With over 23 citations across his key publications, Zhu’s research is recognized for its clarity and practical relevance, offering tools that are directly applicable to robotics, autonomous systems, and sequential decision-making. His work continues to influence the development of data-efficient reinforcement learning methods.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning Policies for Markov Decision Processes From Data

   17 citations · 2018
2. 2
   Learning parametric policies and transition probability models of markov decision processes from data

   6 citations · 2020

Key Collaborators

Contact & Links