Changjie Fan
Papers
1
Total Citations
34
H-Index
1
About
Changjie Fan is a leading researcher in multi-agent reinforcement learning (MARL), with a focus on developing efficient algorithms for sparse-interaction systems—environments where agents only occasionally need to coordinate, as seen in robot swarms or team sports. His most cited work, "Value Function Transfer for Deep Multi-Agent Reinforcement Learning Based on N-Step Returns" (2019, 34 citations), introduces a novel method for reusing single-agent knowledge to accelerate multi-agent learning. By leveraging n-step returns and value function transfer, Fan’s approach significantly reduces the sample complexity and training time in sparse-interaction settings, addressing a critical bottleneck in scaling MARL to real-world applications. This contribution has been influential in advancing transfer learning within multi-agent systems, enabling more practical deployment in domains like autonomous driving and game AI. Fan’s research bridges the gap between single-agent and multi-agent paradigms, offering scalable solutions that have garnered attention from both academia and industry. His work continues to shape how researchers tackle complex, decentralized coordination problems, making him a key figure in the evolution of modern reinforcement learning.
Research Focus
Key Achievements
Top Papers
- 1