Haoqiang Chen

National University of Defense Technology

Papers

Total Citations

H-Index

About

Haoqiang Chen is a researcher whose work lies at the intersection of reinforcement learning (RL) and representation learning, with a particular focus on improving how agents perceive and interact with complex environments. His most cited work, "A2C: Attention-Augmented Contrastive Learning for State Representation Extraction" (2020, 16 citations), tackles a fundamental bottleneck in RL: the quality of state representations. Chen argues that inefficient or poorly structured representations are a root cause of poor learning efficiency and generalization in RL agents. To address this, he introduces a novel framework that augments contrastive learning with attention mechanisms, enabling the agent to extract more salient and compact features from raw sensory data. This approach directly enhances the agent’s ability to focus on task-relevant information, leading to faster convergence and more robust policy learning. By bridging self-supervised representation learning with reinforcement learning, Chen’s work offers a practical pathway toward more sample-efficient and generalizable RL systems—a critical step for deploying RL in real-world applications where data is scarce or environments are dynamic.