Yunhao Tang

About

Yunhao Tang is a researcher at the forefront of reinforcement learning (RL), specializing in the intersection of evolutionary strategies (ES), neural architecture search (NAS), and blackbox optimization. His work addresses fundamental challenges in scaling RL to complex, real-world tasks, particularly in robotics and vision-based environments. Tang’s major contributions include pioneering algorithms like ES-ENAS, which seamlessly integrates ES with efficient NAS to automatically design compact, high-performing RL policies without additional computational cost. This work, along with his research on "Provably Robust Blackbox Optimization for Reinforcement Learning," has garnered significant attention, accumulating over 20 citations across his most-cited papers. Notably, Tang has also advanced vision-based RL by introducing implicit attention mechanisms to unlock pixel-level information, mitigating issues like high dimensionality and observational overfitting. His innovative approaches to combining derivative-free optimization with architecture search have established him as a key figure in making RL more scalable, robust, and practical for deployment in resource-constrained environments. Tang’s research continues to inspire new directions in automated policy design and efficient learning.

Research Focus

Key Achievements

Top Papers

1. 1
   Provably Robust Blackbox Optimization for Reinforcement Learning

   11 citations · 2019
2. 2
   ES-ENAS: Combining Evolution Strategies with Neural Architecture Search at No Extra Cost for Reinforcement Learning.

   9 citations · 2021
3. 3
   Reinforcement Learning with Chromatic Networks for Compact Architecture Search

   2 citations · 2019
4. 4
   Unlocking Pixels for Reinforcement Learning via Implicit Attention

   2 citations · 2021

Key Collaborators

Contact & Links