Thang Nhat Duong
Thang Nhat Duong
Hi, I am a PhD candidate at the University of Arizona, under the instruction of Prof. Chicheng Zhang. Before that, I worked as a Research resident at VinAI from 2019-2022. Besides research, I have three years of working experience as a Junior Software Developer, Team leader, and Scrum Master.
I am interested in how to do efficient planning to solve sequential problems with underline structures. Despite many advances in Reinforcement Learning, there is still a big gap from solving toy problems to useful, practical projects. Since the real world is too high-dimensional, with high signal-to-noise ratio, and the data gathering process may be too costly, I want find a way to discover and exploit the hidden structures of many problems to solve problems more effectively.
What long-term project do you want to work on?
At the moment, I am trying to understand how to balance meta-exploration (discovering hidden structures) with the exploitation objective in the Meta-bandit problem. In the future, I want to better understand how we can apply this hidden knowledge to more general online learning problems. From there, I hope to expand my research to include both self-discover hidden structures and prior knowledge from humans, as well as efficiently update and correct this information.
What do you enjoy most about your work?
I enjoy daydreaming about possible solutions to the problem I am working on. Then, after analyzing and debating with myself about how most of these solutions will not work, I have a better understanding of the problem and a renewed appreciation for state-of-the-art solutions.
What are your career goals?
I hope to see my work becoming useful someday, solving interesting problems, and helping people.
Tell us something interesting about yourself!
After graduating with a B.S.E in Mechatronics Engineering, after two years, I quit my job as a team leader of a Software Developing team to start again as an intern to pursue a research career.