My name is Hong Ye, and I am a 4th and final year PhD student in the Cambridge Image Analysis group, supervised by Prof. Carola-Bibiane Schönlieb. Funded by GSK.ai. I am co-supervised by Prof. Subhadip Mukherjee and Prof. Junqi Tang, and my industrial supervisor is Dr. Alexandra Pettet. I am currently looking for postdoc positions, expected graduation July 2025.

What do I work on?

My current focus is on machine learning theory, such as the manifold hypothesis and links with learning theory, and generally why machine learning works as a function of problem, data, optimizer and network. I aim to answer: what do we need to assume to make machine learning work?

My previous work addresses a similar question: how can we exploit problem structures to make machine learning work? We combine theoretical guarantees with practical applications, developing provable ML-based algorithms using classical mathematics as a skeleton. This includes optimization (convex learning-to-optimize, Plug-and-Play, inverse problems), MCMC sampling (noise-free methods, unsupervised learning/ imaging), and differential geometry (manifold hypothesis, intrinsic dataset complexity).

Other tidbits

• I was the youngest member admitted to the University of Hong Kong in 2015 (aged 11). I am also probably the youngest member (aged 13) admitted to Cambridge in recent history, but I have not verified this. I am now 21, if you are interested.

• I do all the coding and 98% of the writing for my first-author papers. Feel free to use the code from any of my work, and do send me an email if you have any questions, as the codebases online are quite dirty due to scattered computational resources.

• I am a Masason Foundation scholar of 6 years.

• I also like classical piano, food, badminton, and birds. I have also dabbled in fencing and small-bore rifle shooting.