Yutong Wang

PhD Candidate | University of Michigan | yutongw@umich.edu

Hi! I am a sixth year PhD candidate in Electrical Engineering and Computer Science. My research is in machine learning. I am advised by Clay Scott.

Research interests

I’m motivated by the following questions:

• How do (partially) quantized neural networks compare to their full-precision counterparts, from the point-of-view of optimization, function approximation, and generalization?
• Can we close the gap between multiclass and binary SVMs? (In terms of theory and availability of solvers)

Publications

• VC dimension of partially quantized neural networks in the overparametrized regime
  Yutong Wang and Clayton Scott
  [arXiv] [Code]
• An exact solver for the Weston-Watkins SVM subproblem
  Yutong Wang and Clayton Scott
  ICML 2021.
  [Paper] [arXiv] [Code]
• Weston-Watkins Hinge Loss and Ordered Partitions
  Yutong Wang and Clayton Scott
  NeurIPS 2020.
  [Paper] [arXiv]
• Hybrid Stem Cell States: Insights Into the Relationship Between Mammary Development and Breast Cancer Using Single-Cell Transcriptomics
  Tasha Thong, Yutong Wang, Michael D. Brooks, Christopher T. Lee, Clayton Scott, Laura Balzano, Max S. Wicha, and Justin A. Colacino
  Frontiers in Cell and Developmental Biology.
  [Paper] [Supporting technical report]

Blog posts

• 2020/12/30 Kernel multiclass support vector machine solvers
• 2020/11/06 The Weston-Watkins SVM dual problem
• 2020/10/14 SVM solver jargons
• 2020/10/02 A non-exhaustive list of SVM solvers