About Myself

I am a Quantitative Researcher at Citadel Securities. Previously, I got my Ph.D. degree from Stanford ICME, where I was fortunate to be advised by Professor Tengyu Ma. My research interest lies in machine learning theory, in particular Federated Learning, Optimization and Deep Learning theory. Prior to attending Stanford, I earned dual B.S. degrees in Mathematics and Computer Science from Peking University.

Please find my CV and Google Scholar here.

Publications

• Plex: Towards Reliability using Pretrained Large Model Extensions
  with Dustin Tran et al.
  arXiv: 2207.07411 | bibtex | code | blog

• On Principled Local Optimization Methods for Federated Learning
  Ph.D. Thesis, Stanford University, 2022

• Big-Step-Little-Step: Efficient Gradient Methods for Objectives with Multiple Scales
  with Jonathan Kelner, Annie Marsden, Vatsal Sharan, Aaron Sidford, Gregory Valiant (alphabetical order)
  COLT 2022 | bibtex | code

• Sharp Bounds for Federated Averaging (Local SGD) and Continuous Perspective
  Margalit Glasgow*, Honglin Yuan*, Tengyu Ma (*equal contribution)
  AISTATS 2022 | bibtex | code

• What Do We Mean by Generalization in Federated Learning?
  Honglin Yuan, Warren Morningstar, Lin Ning, Karan Singhal
  ICLR 2022 | bibtex | code

• A Field Guide to Federated Optimization
  with Jianyu Wang et al.
  arXiv:2107.06917 | bibtex | code

• Federated Composite Optimization
  Honglin Yuan, Manzil Zaheer, Sashank Reddi
  ICML 2021 | bibtex | talk video at FLOW seminar | slides | poster | code

• Federated Accelerated Stochastic Gradient Descent
  Honglin Yuan, Tengyu Ma
  NeurIPS 2020 (Best Paper in FL-ICML’20 workshop) | bibtex | video (3 min) | poster | code

• Global Optimization with Orthogonality Constraints via Stochastic Diffusion on Manifold
  Honglin Yuan, Xiaoyi Gu, Rongjie Lai, Zaiwen Wen
  Journal of Scientific Computing, 2019 | bibtex | slides