Deheng Ye

About

Deheng Ye is a leading researcher at the intersection of deep reinforcement learning (RL) and continual learning, with a focus on building adaptive, lifelong learning agents. His work addresses a fundamental challenge in modern AI: how to move beyond tabula rasa learning to create systems that can continuously adapt to changing environments without catastrophic forgetting. In his highly cited survey, "Pretraining in Deep Reinforcement Learning: A Survey" (2022, 11 citations), Ye provided a comprehensive roadmap of the rapidly evolving field, synthesizing breakthroughs from games to robotics and establishing a framework for leveraging prior knowledge in RL. His pioneering contribution, "Dynamics-Adaptive Continual Reinforcement Learning via Progressive Contextualization" (2023, 9 citations), introduced a novel method for enabling RL agents to dynamically adapt their behavior as environments shift over their lifetime. This work directly tackles the dual challenge of prompt adaptation and memory retention, offering a principled approach to progressive contextualization. Ye's research is shaping the next generation of AI systems that can learn continuously and robustly in real-world, non-stationary environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Pretraining in Deep Reinforcement Learning: A Survey

   11 citations · 2022
2. 2
   Dynamics-Adaptive Continual Reinforcement Learning via Progressive Contextualization

   9 citations · 2023

Key Collaborators

Contact & Links