Yunfu Deng
Papers
1
Total Citations
2
H-Index
1
About
Yunfu Deng is a researcher advancing reinforcement learning (RL), with a primary focus on tackling the challenge of sparse reward environments. His most notable contribution is the development of USHER (Unbiased Sampling for Hindsight Experience Replay), a 2022 work that refines the widely used Hindsight Experience Replay (HER) algorithm. While HER cleverly reuses failed trajectories as successful ones for different goals, it can introduce bias into the learning process. Deng’s USHER method corrects this by implementing an unbiased sampling mechanism, enabling more stable and efficient policy learning in complex, goal-oriented tasks. This work has garnered early citations, signaling its growing influence in the RL community. By addressing a fundamental limitation of a popular technique, Deng’s research helps bridge the gap between simulated RL and real-world applications, such as robotics and autonomous systems, where reward signals are often sparse and delayed. His contributions are paving the way for more robust and sample-efficient reinforcement learning algorithms.
Research Focus
Key Achievements
Top Papers
- 1USHER: Unbiased Sampling for Hindsight Experience Replay2 citations · 2022