Hong Ye Tan

Hong Ye Tan is currently a Hedrick Assistant Adjunct Professor in Computational and Applied Mathematics at the University of California, Los Angeles (UCLA), hosted by Professor Stanley Osher. Previously, he completed his PhD at the University of Cambridge's Department of Applied Mathematics and Theoretical Physics as a member of the Cambridge Image Analysis group and the Cantab Capital Institute for the Mathematics of Information, supervised by Professors Carola-Bibiane Schönlieb, Subhadip Mukherjee, and Junqi Tang with funding from GSK.ai.

His educational trajectory is exceptional: admitted to the University of Hong Kong at age 11 in 2015 (youngest in recent history) and to Cambridge at age 13 for doctoral studies. He passed his PhD thesis with no corrections, focusing on provably convergent algorithms leveraging geometric structures in data.

Tan's research centers on machine learning theory, specifically investigating why learning succeeds through interactions between problem structure, data distributions, optimizers, and network architectures. His work bridges differential geometry (manifold hypothesis, intrinsic complexity), optimization (convex learning-to-optimize, Plug-and-Play inverse problems), and sampling theory (noise-free MCMC methods). He develops theoretically grounded algorithms with practical applications in imaging and unsupervised learning, emphasizing provable convergence guarantees derived from classical mathematics.

Analysis of his 13 recent publications reveals a cohesive research program connecting optimal transport theory, manifold learning, and regularization techniques. His work demonstrates how geometric insights enable efficient solutions for high-dimensional problems, particularly in image analysis where dimensionality effects transform from curse to blessing. Key themes include Wasserstein proximal methods, dataset distillation via quantization, and accelerating mirror descent through equivariance.

His scientific recognition includes:

• Masason Foundation Fellowship
• GSK.ai PhD Fellowship

Tan has secured research funding through the GSK.ai PhD studentship and operates within Professor Stanley Osher's group at UCLA. He maintains active collaborations from his Cambridge tenure, particularly with the Cambridge Image Analysis group. Notably, he handles 100% of coding and 98% of writing for first-author publications, actively encouraging code reuse by the community. His work continues to explore foundational questions in learning theory while developing practical tools for inverse problems and imaging science.