Dong-Sheng Ding

Papers

Total Citations

H-Index

About

Dong-Sheng Ding is a rising researcher in reinforcement learning and multi-agent systems, whose work bridges optimization theory and practical decision-making in complex environments. His research focuses on developing efficient algorithms for multi-agent temporal-difference learning and constrained Markov decision processes (MDPs), particularly in continuous state and action spaces. Ding’s most-cited paper, “Fast Multi-Agent Temporal-Difference Learning via Homotopy Stochastic Primal-Dual Optimization” (2019, 12 citations), introduces a novel homotopy-based primal-dual framework that accelerates policy evaluation in multi-agent settings, enabling agents with local observations to collaboratively learn value functions through communication over connected networks. His more recent work, “Deterministic Policy Gradient Primal-Dual Methods for Continuous-Space Constrained MDPs” (2025, 2 citations), addresses the challenge of computing deterministic optimal policies under constraints—a critical problem in robotics, autonomous systems, and resource allocation. By combining primal-dual methods with deterministic policy gradients, Ding provides a principled approach to handling safety and resource constraints in continuous control tasks. His contributions are particularly valuable for students and researchers working on scalable multi-agent learning, constrained optimization, and real-world reinforcement learning applications where theoretical guarantees and computational efficiency are paramount.