I am a PhD student in the Statistics Department at UCLA. I work on mathematical theory for deep learning.
I am fortunate to be advised by Guido Montúfar.
liangshuang at g.ucla.edu / Google scholar / Short CV
My current research focuses on optimization in neural networks. Specifically, I aim to better understand:
The optimization trajectory in parameter space;
The implicit bias of the optimization algorithm (which model the algorithm tends to select);
How these aspects are influenced by network architecture, optimizer, initialization, step size, etc.
Previously, I studied persistent homology, a data representation method in topological data analysis.
Implicit Bias of Mirror Flow for Shallow Neural Networks in Univariate Regression.
Shuang Liang, Guido Montúfar. ICLR 2025 (Spotlight). Preprint [arXiv:2410.03988]. Virtual poster [SlidesLive].
Pull-back Geometry of Persistent Homology Encodings.
Shuang Liang, Renata Turkeš, Jiayi Li, Nina Otter, Guido Montúfar. TMLR 2024. Preprint [arXiv:2310.07073]. Repo [GitHub]. Video [Video].