Publications

Compositional Behavioral Semantics for State Abstraction in Reinforcement Learning

State abstraction plays a key role in scaling reinforcement learning to complex but structured systems. In studying such systems, a …
PDF Poster
Compositional Behavioral Semantics for State Abstraction in Reinforcement Learning

Understanding Behavioral Metric Learning

A large-scale study on the impact of behavioral metric learning in deep RL, conceptually unifying and evaluating recent methods spanning 370 tasks with diverse noise settings.
PDF Code Poster Slides DOI
Understanding Behavioral Metric Learning

Discovering Temporal Structure: An Overview of Hierarchical Reinforcement Learning

Click to view the blog. An 80+ page comprehensive overview of HRL, focusing on methods for discovering temporal structure and the key benefits it provides.
PDF Slides DOI
Discovering Temporal Structure: An Overview of Hierarchical Reinforcement Learning

Chronosymbolic Learning

A novel framework for efficient Constrained Horn Clause (CHC) solving, “Chronosymbolic Learning”, which provides fundamentals to unify symbolic-reasoning-based and data-driven approaches synergistically.
PDF Code Poster Slides DOI
Chronosymbolic Learning

RLMob - Deep Reinforcement Learning for Successive Mobility Prediction

Introduce RL to human mobility prediction to address the key challenges in this task.
PDF Code Slides Video DOI
RLMob - Deep Reinforcement Learning for Successive Mobility Prediction

Match Plan Generation in Web Search with Parameterized Action Reinforcement Learning

During my Microsoft internship, we explored an RL-based prototype for match plan generation, which outperformed hand-crafted match plans tuned by experts for years, and was later integrated into Bing.
PDF Code Slides Video DOI
Match Plan Generation in Web Search with Parameterized Action Reinforcement Learning