Haichuan Yang
Papers
1
Total Citations
5
H-Index
1
About
Haichuan Yang is a researcher whose work lies at the intersection of reinforcement learning and continuous control, with a particular focus on developing robust and efficient algorithms for complex decision-making domains. His most notable contribution is the introduction of **Marginal Policy Gradients (MPG)** , a unified family of estimators designed to handle bounded action spaces—a common challenge in robotics and real-time strategy games. By providing a principled framework that bridges discrete and continuous action parameterizations, Yang’s work offers a more stable and sample-efficient alternative to traditional policy gradient methods. While his highly cited paper has garnered 5 citations, the conceptual depth and practical relevance of his approach have influenced subsequent research in safe and constrained reinforcement learning. Yang’s research is particularly valuable for students and practitioners working on real-world control problems where action bounds are inherent, such as autonomous navigation or game AI. His contributions underscore the importance of tailoring gradient estimation to the structure of the action space, a key insight for advancing scalable and reliable learning algorithms.
Research Focus
Key Achievements
Top Papers
- 1