Fangshuo Liao šŸ’­
Fangshuo Liao Jasper

Ph.D. Candidate

Rice University, Computer Science Department

About Me

My name is Fangshuo Liao (å»–čŠ³ēƒ), and my preferred name is Jasper. I am a PhD candidate in Computer Science at Rice University in Houston, TX, advised by Dr. Anastasios Kyrillidis. I also participate in the AI-OWLS research cluster. Before my PhD, I received a Bachelor of Science in Computer Science and Bachelor of Arts in Mathematics from Rice University in 2020.

My research interest lies broadly in deep learning theory, convex and nonconvex optimization, as well as developing methods that provably improve the efficiency of solving machine learning problems.

Aside from work, I am interested in cooking, coffee making, board games, and murder mystery games (å‰§ęœ¬ę€). In the past two years, I also participated in the theatrical productions at Rice Chinese Theater Club, where I was the president, production stage manager, and an actor.

I have three cats, and I love them a lot!

Download CV
Interests
  • Convex/Nonconvex Optimization
  • Deep Learning Theory
  • Efficient ML Algorithms
  • Low Dimensional Models/Pruning
Education

  • Ph.D. Computer Science

    Rice University

  • B.S. Computer Science

    Rice University

  • B.A. Mathematics

    Rice University

šŸ“š My Research

Consider Algorithm A for Problem B that naturally arises in the broad field of machine learning. I try to answer the following questions:

  • What is the theoretical guarantee of Algorithm A?
  • What are the mathematical properties of Problem B?
  • How to leverage the mathematical properties of Problem B to design a better version of Algorithm A?

My recent focus is on the theory of deep learning.

Featured Publications
Provable Model-Parallel Distributed Principal Component Analysis with Parallel Deflation
Provable Accelerated Convergence of Nesterovā€™s Momentum for Deep ReLU Neural Networks
Recent Publications
Fangshuo Liao, Wenyi Su, Anastasios Kyrillidis (2025). Provable Model-Parallel Distributed Principal Component Analysis with Parallel Deflation. CPAL.
PDF Cite Code Poster
Fangshuo Liao, J. Lyle Kim, Cruz Barnum, Anastasios Kyrillidis (2024). On the Error-Propagation of Inexact Hotelling's Deflation for Principal Component Analysis. ICML.
PDF Cite Poster
Fangshuo Liao, Anastasios Kyrillidis (2024). Provable Accelerated Convergence of Nesterovā€™s Momentum for Deep ReLU Neural Networks. ALT.
PDF Cite Poster
Cameron R. Wolfe*, Fangshuo Liao*, Qihan Wang, J. Lyle Kim, Anastasios Kyrillidis (2024). How Much Pre-training Is Enough to Discover a Good Subnetwork?. TMLR.
PDF Cite Code
Zichang Liu, Aditya Desai, Fangshuo Liao, Weitao Wang, Victor Xie, Zhaozhuo Xu, Anastasios Kyrillidis, Anshumali Shrivastava (2023). Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Time. NeurIPS.
PDF Cite
See all publications
Recent & Upcoming Talks
Provable Accelerated Convergence of Nesterovā€™s Momentum for Deep ReLU Neural Networks
Recent News