"Ryan" Zhiyu An

Hi! I'm a PhD student at University of California, Merced, advised by Professor Wan Du. Previously, I completed a BA at UC Berkeley in 1.5 years and started my PhD at age 19.

My work primarily contributes to the safe reinforcement learning community. I improve the trustworthiness of black-box AI models for real-world control and decision-making tasks. I use techniques such as regularized optimization, formal methods, and uncertainty estimation to steer AI agents' behaviors, interpret black-box models, and manage risks from out-of-distribution scenarios; all to make AI safer for users.

As the first author, I have published in DAC, ICLR, BuildSys, and SenSys. I was named Best Paper Award Runner-Up at BuildSys 2023 for my contributions to safe reinforcement learning for HVAC control. As a co-author, I have published in KDD. I have served as a reviewer for ACM SenSys, IEEE IOTJ, T-NSE, etc.

I'm open to job / research / startup opportunities. Check out the rest of this site or CV to learn more, and feel free to reach out!


Email     G. Scholar     LinkedIn     CV    

                           
      

2021

Beijing Aidi High School - Scholarship ($ 20k+)

2020

Beijing Aidi High School - Scholarship ($ 20k+)

2019

Beijing Aidi High School - Scholarship ($ 20k+)

Australian Science Olympiad in Chemistry - 🏅 Gold Medal

2018

Australian Science Olympiad in Chemistry - 🏅 Gold Medal

2017

Australian Mathematics Competition - High Distinction (Top 3%)

Golden-Retriever: High-Fidelity Agentic Retrieval Augmented Generation for Industrial Knowledge Base

Zhiyu An, Xianzhong Ding, Yen-Chun Fu, Cheng-Chung Chu, Yan Li, Wan Du
arXiv 2024
Paper  •   bibtex

MARLP: Time-series Forecasting Control for Agricultural Managed Aquifer Recharge

Yuning Chen, Kang Yang, Zhiyu An, Brady Holder, Luke Paloutzian, Khaled M. Bali, Wan Du
Knowledge Discovery and Data Mining (KDD) 2024
Paper  •   Code  •   Slides  •   bibtex

Go Beyond Black-box Policies: Rethinking the Design of Learning Agent for Interpretable and Verifiable HVAC Control

Zhiyu An, Xianzhong Ding, Wan Du
Design Automation Conference (DAC) 2024
Paper  •   Code  •   Slides  •   Presentation Recording  •   Poster  •   bibtex

Reward Bound for Behavioral Guarantee of Model-based Planning Agents

Zhiyu An, Xianzhong Ding, Wan Du
Tiny Paper Track, International Conference on Learning Representations (ICLR) 2024
Paper  •   OpenReview  •   bibtex

CLUE: Safe Model-Based Reinforcement Learning HVAC Control Using Epistemic Uncertainty Estimation

Zhiyu An, Xianzhong Ding, Arya Rathee, Wan Du
The ACM International Conference on Energy-Efficient Buildings, Cities, and Transportation (BuildSys) 2023★ Best Paper Award Runner-Up ★
Paper  •   Code  •   Slides (8min)  •   Slides (30min)  •   bibtex

Data Efficient HVAC Control using Gaussian Process-based Reinforcement Learning

Zhiyu An, Xianzhong Ding, Wan Du
The ACM Conference on Embedded Networked Sensor Systems (SenSys), Poster 2023
Poster Abstract  •   Poster

2024

Program Committee Member, Trustworthy Machine Learning Workshop at CIKM

Reviewer, IEEE Internet of Things Journal (IOTJ)

Reviewer, IEEE Transactions on Network Science and Engineering (T-NSE)

Reviewer, Data Centric Engineering, Cambridge University Press

2023

Reviewer, ACM SenSys

Speaker Selection Committee Member, TEDxUCMerced

Judge, NASA Space Apps Challenge Hackathon