Dezhong Peng

Papers

Total Citations

H-Index

About

Dezhong Peng is a researcher whose work bridges computer vision and reinforcement learning, with a particular focus on advancing image retrieval and policy optimization. His most cited paper, "Relaxed Energy Preserving Hashing for Image Retrieval" (2024, 19 citations), tackles a critical challenge in industrial robotics: enabling efficient and accurate visual search for applications like machine vision, street view search, and object grasping. By developing a novel learning-to-hash method that relaxes energy constraints, Peng improves the speed and precision of image retrieval systems, directly impacting real-world robotic perception. In reinforcement learning, his work "Proximal Policy Optimization with Future Rewards" (2021, 3 citations) addresses the instability of gradient estimation in policy gradient algorithms, enhancing the widely-used PPO framework by incorporating future reward signals. This contribution demonstrates Peng’s ability to refine foundational algorithms for more stable and effective learning. With a growing citation record and a focus on practical, high-impact problems, Dezhong Peng is establishing himself as a thoughtful contributor to both visual computing and intelligent decision-making systems.