Yanchao Sun (孙妍超)

Email: ycs at umd dot edu
Google Scholar
Github
LinkedIn

About me

I am a Ph.D. candidate in University of Maryland, College Park, where I am fortunate to be advised by Prof. Furong Huang. My research mainly focuses on reinforcement learning (RL), including adversarial RL, multi-task RL, sample-efficient RL, etc. My long-term goal is to make RL agents more robust and more efficient. Before that, I received my Bachelor’s degree in Computer Science from Sichuan University, where I closely worked with Prof. Ning Yang on data mining and recommender systems.

In the past summer, I was excited to work as a research intern in Microsoft with Dr. Shuang Ma, working on pretraining methods for RL. I also had wonderful experience working as a research intern in the AI Research team of JPMorgan Chase & Co. under the supervision of Dr. Sumitra Ganesh; We worked on certifiable defenses against communication attacks in Multi-agent RL (paper). In the summer of 2020, I interned at Unity Technologies with my mentor Dr. Andrew Cohen; We investigated knowledge transfer across tasks (paper).

I am looking for full-time opportunities. Here are my CV and my Research Statement. Please feel free to contact me if you think I could be a good fit.

News and Highlights

    [2023.1]    We have 3 papers (1 spotlight, 2 posters) accepted by ICLR 2023!
    [2022.11]    I am co-organizing the first workshop on Reincarnating RL in ICLR 2023. Check out our website: reincarnating-rl.github.io.
    [2022.09]    We have 3 papers accepted by NeurIPS 2022. Many thanks to my collaborators!
    [2022.01]    Two of our papers have been accepted by ICLR 2022. Please check out our papers on adversarial RL and transfer RL.
    [2021.12]    Our paper "Who Is the Strongest Enemy? Towards Optimal and Efficient Evasion Attacks in Deep RL" was honored to get the Best Paper Award of the SafeRL 2021 Workshop!
    [2021.12]    Presented our papers in the NeurIPS 2021 DeepRL and SafeRL workshops.
    [2021.06]    Started a research internship at JPMorgan Chase & Co., under the supervision of Dr. Sumitra Ganesh.
    [2021.01]    Our paper on poisoning RL was accepted by ICLR 2021.

Selected Publications

Please see here for a full list of my publications.

ICLR 2023          

SMART: Self-supervised Multi-task pretrAining with contRol Transformers.

Yanchao Sun, Shuang Ma, Ratnesh Madaan, Rogerio Bonatti, Furong Huang, Ashish Kapoor
In International Conference on Learning Representations, 2023

Paper (Code and models coming soon!)

NeurIPS 2022          

Efficient Adversarial Training without Attacking: Worst-Case-Aware Robust Reinforcement Learning.

Yongyuan Liang*, Yanchao Sun*, Ruijie Zheng, and Furong Huang. (* Equal Contribution)
In the 36th Conference on Neural Information Processing Systems, 2022

Paper     Code

ICLR 2022          

Who Is the Strongest Enemy? Towards Optimal and Efficient Evasion Attacks in Deep RL.

Yanchao Sun, Ruijie Zheng, Yongyuan Liang, and Furong Huang.
In International Conference on Learning Representations, 2022
Best Paper Award in NeurIPS 2021 Workshop of Safe and Robust Control of Uncertain Systems (SafeRL 2021).

Paper     Code     HTML

ICLR 2022          

Transfer RL across Observation Feature Spaces via Model-Based Regularization.

Yanchao Sun, Ruijie Zheng, Xiyao Wang, Andrew Cohen, and Furong Huang.
In International Conference on Learning Representations, 2022

Paper     Code     HTML

ICLR 2021          

Vulnerability-Aware Poisoning Mechanism for Online RL with Unknown Dynamics.

Yanchao Sun, Da Huo, and Furong Huang.
In International Conference on Learning Representations, 2021

Paper     Code     HTML

AAAI 2021          

TempLe: Learning Template of Transitions for Sample Efficient Multi-task RL.

Yanchao Sun, Xiangyu Yin, and Furong Huang.
In AAAI Conference on Artificial Intelligence, 2021

Paper     Code     HTML

Please see here for a full list of my publications.