David Yu-Tung Hui / 許宇同

There are multiple ways to write my name. In Latin script, my surname is "Hui" and my firstname is "David Yu-Tung." In Traditional Chinese characters, my family name is "許" and my given name is "宇同." Most people call me "David." Others call me "宇同" or "Yu-Tung."

I am currently unemployed. I used to be an AI researcher in deep reinforcement learning. I wrote two works improving the optimization stability of off-policy gradient-based Q-learning algorithms.

1. Stabilizing Q-Learning for Continuous Control
   David Yu-Tung Hui
   MSc Thesis, University of Montreal, 2022
   I derived a deep reinforcement learning algorithm from mathematical first principles. I derived the SACLite loss functions from the principle of maximum-entropy and justified the use of LayerNorm with a neural-tangent-kernel-inspired analysis. Compared to baseline actor-critic algorithms, my algorithm did not diverge in high-dimensional continuous control.
   [.pdf] [Errata]

2. Double Gumbel Q-Learning
   David Yu-Tung Hui, Aaron Courville, Pierre-Luc Bacon
   Spotlight at NeurIPS 2023
   We showed that Q-learning with function approximation has two previously unnoticed heteroscedastic Gumbel noise sources. An algorithm accounting for these noise sources attained almost 2 times the aggregate asymptotic performance of the popular SAC baseline.
   [.pdf] [Reviews] [Poster (.png)] [5-min talk] [1-hour seminar] [Code (GitHub)] [Errata]

The best way to contact me is email. My email address is listed in one of my written works.