Yat Long Lo
Papers
2
Total Citations
16
H-Index
2
About
Yat Long Lo is a researcher whose work focuses on the intersection of representation learning and reinforcement learning, with a particular emphasis on how neural network generalization impacts agent performance. His most-cited paper, "Improving Performance in Reinforcement Learning by Breaking Generalization in Neural Networks" (2020, 8 citations), makes a counterintuitive yet impactful contribution: it challenges the conventional wisdom that strong generalization is always beneficial in RL. Lo demonstrates that deliberately breaking certain forms of generalization in neural network representations can actually improve learning efficiency and task performance, especially in complex, high-dimensional domains. This insight offers a fresh perspective on the design of deep RL systems, which have historically struggled with scalability and domain-specific tuning. By questioning a core assumption in deep learning, Lo’s work opens new avenues for building more robust and adaptable agents. His research is particularly valuable for students and practitioners seeking to understand the nuanced relationship between representation quality and reinforcement learning success, highlighting that sometimes, less generalization can lead to more effective learning.
Research Focus
Key Achievements
Top Papers
- 1
- 2