About
I am currently an Institute for Foundations of Data Science (IFDS) Postdoctoral Fellow at the University of Washington, working with Dmitriy Drusvyatskiy and Maryam Fazel. I received my PhD in Computer Science from UCSD, where I was fortunate to be advised by Mikhail Belkin. Prior to UCSD, I received my BS in Mathematics from Zhejiang University.
I am broadly interested in the optimization and mathematical foundations of machine learning. My core research interest is in establishing the mathematical foundations for modern deep learning and designing theoretically grounded algorithms that advance existing methods. In recent work, we showed that kernel methods are capable of implementing feature and hierarchical learning (paper).
Email: libinzhu at uw dot edu
🚀 I am currently on the job market for full‑time positions.
Selected Publications
View All →Iteratively reweighted kernel machines efficiently learn sparse functions
Libin Zhu, Damek Davis, Dmitriy Drusvyatskiy, Maryam Fazel
arXiv:2505.08277
Emergence in non-neural models: grokking modular arithmetic via average gradient outer product
Neil Rohit Mallinar, Daniel Beaglehole, Libin Zhu, Adityanarayanan Radhakrishnan, Parthe Pandit, Mikhail Belkin
ICLR (Oral)
Assembly and iteration: transition to linearity of wide neural networks
Chaoyue Liu, Libin Zhu, Mikhail Belkin
Applied and Computational Harmonic Analysis
Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning
Libin Zhu, Chaoyue Liu, Adityanarayanan Radhakrishnan, Mikhail Belkin
ICML
Loss landscapes and optimization in over-parameterized non-linear systems and neural networks
Chaoyue Liu, Libin Zhu, Mikhail Belkin
Applied and Computational Harmonic Analysis
On the linearity of large non-linear models: when and why the tangent kernel is constant
Chaoyue Liu, Libin Zhu, Misha Belkin
NeurIPS(Spotlight)