Welcome to my website!

Short Bio 

I am an incoming Assistant Professor at National Taiwan University (Electrical Engineering) and currently a Postdoctoral Fellow in the School of Electrical and Computer Engineering at Georgia Institute of Technology, where I am very fortunate to work with Prof. Pan Li. I received my Ph.D. degree in Electrical and Computer Engineering at the University of Illinois Urbana - Champaign. I am very fortunate to work with my advisor Prof. Olgica Milenkovic. Before that, I got my B.S. degree in Electrical Engineering from National Taiwan University where I worked with Prof. I-Hsiang Wang.  

My research interests lie in the field of Regulatable AI, in particular privacy and graph-related machine learning. I currently focus on machine unlearning, differential privacy, AI copyright infringement, and their application to graph ML. My other works include geometric learning, including general (hyper)graph neural networks, active learning on (hyper)graphs, and classification in hyperbolic/product spaces.

I spent two summers in Amazon Search and worked with Wei-Cheng Chang, Cho-Jui Hsieh, Jiong Zhang, Jyun-Yu Jiang,  Hsiang-fu Yu, and Inderjit Dhillon. Our works are about extracting high-quality node features from raw data with the help of graph information (paper accepted by ICLR 2022) and improving eXtreme multilabel classification methods via side-information with graph learning techniques (paper accepted by ICML 2023). I spent a summer in Nokia Bell Labs and worked with Antonia Maria Tulino and Jaime Llorca. Our work is about active learning on the geometric block model (paper accepted by AAAI 2019).

I've been invited (previously or currently) as an area chair of NeurIPS, and a reviewer of NeurIPS, ICML, ICLR, AISTATS, KDD, TMLR, TheWebConf, SDM, ISIT, CoLLAs, TPDP.

News

• 2025/04: I will serve as an area chair for NeurIPS 2025!

• 2025/03: I will join National Taiwan University as an Assistant Professor starting in February 2026! I really appreciate everyone who has helped me during my job search journey!

• 2025/03: I will give a talk at School of CSE Seminar Series at GaTech on the topic: "Machine Unlearning: The General Theory and LLM Practice of Privacy"!

• 2025/01: I will serve as a senior reviewer for CoLLAs 2025!

• 2025/01: Two papers accepted by ICLR 2025!

  • Convergent privacy loss of noisy-sgd without convexity and smoothness: This is the one I lead, collaborating with Pan. I really appreciate the valuable comments provided by reviewers, AC, and my friends.

  • LayerDAG: A Layerwise Autoregressive Diffusion Model for Directed Acyclic Graph Generation (Spotlight): This one is led by Mufei. Big congrats to him and all the collaborators! 

• 2024/11: I gave a talk at the National Taiwan University EE department on the topic: Machine Unlearning: the General Theory and LLM Practice for Privacy! Really nice to be back to NTUEE, reminds me of many good old times.

• 2024/09: Three papers about privacy theory (Unlearning, Graph DP) accepted by NeurIPS 2024! Thanks to all my collaborators and congrats to Rongzhe for leading the DP PageRank paper!

  • Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning (Spotlight)

  • Certified Machine Unlearning via Noisy Stochastic Gradient Descent

  • Differentially Private Graph Diffusion with Applications in Personalized PageRanks

• 2024/09: Our paper "On the Inherent Privacy Properties of Discrete Denoising Diffusion Models" is accepted by TMLR. Congrats to Rongzhe for leading the project!

• 2024/09: I will give an invited talk about machine unlearning at TAMU hosted by Kuan-Hao Huang in October.

• 2024/07: I will present our work "Langevin Unlearning" at TPDP 2024 in August!

• 2024/07: Our survey paper "Exploring the Opportunities and Challenges of Graph Neural Networks in Electrical Engineering" was accepted by Nature Reviews Electrical Engineering. Big congrats to the whole team!

• 2024/07: I'm excited to share the news that Prof. Pan Li and I will organize two mini-symposiums at MDS 2024, which will be held in Atlanta around October 21 – 25 this year! Please check out the detailed information below, where we are happy to invite many domain experts as our speakers! Please join us if you are interested in Privacy or Graph Machine Learning!

  • • New Frontier of Privacy in Machine Learning

Speakers: Chuan Guo (Meta), Kamalika Chaudhuri (UCSD), Raman Arora (JHU), Wanrong Zhang (Harvard)

• • • New Frontier of Graph Machine Learning

Speakers: Santiago Segarra (Rice), Ziang Chen (MIT), Hanghang Tong (UIUC), Jiliang Tang (MSU)

• 2024/05: Our paper "Machine Unlearning of Pre-trained Large Language Models" is accepted by ACL 2024 as the main paper [arXiv, code]! Kudos to Jin and Xiang for leading this project!

• 2024/04: I gave a talk at Google about "Langevin Unlearning".

• 2024/04: Our project "Privacy-preserving Machine Learning on Graphs" got funded! Many thanks to NSF! I really learned a lot from Prof. Pan Li, Prof. Olgcia Milenkovic, and Prof. Kamalika Chaudhuri during this process, and look forward to future collaborations. 

• 2024/04: Our paper "Langevin Unlearning" is accepted by the ICLR 2024 PrivML workshop as the spotlight presentation! [arXiv (long version)]. Feel free to also check out "Stochastic Gradient Langevin Unlearning" for improved results under the convexity assumption [arXiv].

• 2024/03: I gave an invited talk about machine unlearning at CISS 2024 at Princeton.

• 2024/02: I gave a talk "Machine Unlearning: Current Challenges and Beyond" at the CSIP Seminar at GaTech.

• 2024/01: Our paper "Breaking the Trilemma of Privacy, Utility, Efficiency via Controllable Machine Unlearning" is accepted by TheWebConf 2024 [arXiv]. Congrats to Frank and Gorden who lead this project!

• 2024/01: Our paper "Federated Classification in Hyperbolic Spaces via Secure Aggregation of Convex Hulls" is accepted by TMLR [TMLR]! Congrats to all collaborators Saurav, Jin, Chao, and Professor Milenkovic.

• 2023/09: Our paper "Differentially Private Decoupled Graph Convolutions for Multigranular Topology Protection" is accepted by NeurIPS 2023 [arXiv, code]!

Older updates can be found here.

Affiliations