Yinan Huang

Ph.D. student, Machine Learning, Georgia Institute of Technology Google Scholar E-mail: yhuang903@gatech.edu

About

I am a fourth-year Ph.D. student in Machine Learning at Georgia Tech, supervised by Prof. Pan Li. I am broadly interested in AI for science (developing data-driven methods for accelerating scientific discovery) and Science of AI (understanding the foundations and principles that govern modern AI systems). My current research is focused on

• geometric deep learning: learning on complex data and systems such as graphs.

• generative models: diffusion and flow models, and their applications in spatio-temporal data modeling.

Prior to that, I earned my M.S. degree in ECE at Duke University and B.S. degree in Physics at Sun Yat-sen University. I was a research intern at NEC Lab America working with Martin Min, and a research intern at Peking University working with Prof. Muhan Zhang.

News

2026/3 Our paper about graph SSM is accepted by TPAMI.

2026/2 Our paper Powers of Magnetic Graph Matrix: Fourier Spectrum, Walk Compression, and Applications is accepted by PNAS! A very interesting work studying the connection between random walk on directed networks and the Fourier transform of the Magnetic Laplacian.

Selected Publications

• Accelerated Sequential Flow Matching: A Bayesian Filtering Perspective
  Yinan Huang, Hans Hao-Hsun Hsu, Junran Wang, Bo Dai, Pan Li
  ICLR 2026 ReALM–GEN Workshop (Spotlight)

• Powers of Magnetic Graph Matrix: Fourier Spectrum, Walk Compression, and Applications
  Yinan Huang, David F. Gleich, Pan Li
  The Proceedings of the National Academy of Sciences (PNAS) 2026

• What Are Good Positional Encodings for Directed Graphs?
  Yinan Huang, Haoyu Wang, Pan Li.
  ICLR 2025

• On the Stability of Expressive Positional Encodings for Graphs
  Yinan Huang*, William Lu*, Joshua Robinson, Yu Yang, Muhan Zhang, Stefanie Jegelka, Pan Li
  ICLR 2024

• Boosting the Cycle Counting Power of Graph Neural Networks with I2-GNNs
  Yinan Huang, Xingang Peng, Jianzhu Ma, Muhan Zhang
  ICLR 2023

• 3DLinker: An E(3) Equivariant Variational Autoencoder for Molecular Linker Design
  Yinan Huang, Xingang Peng, Jianzhu Ma, Muhan Zhang
  ICML 2022 (Oral).