3 papers
cs.LG2025
Position: Many generalization measures for deep learning are fragile
Shuofeng Zhang, Ard Louis
In this position paper, we argue that many post-mortem generalization measures -- those computed on trained networks -- are \textbf{fragile}: small training modifications that bare…
cs.LG2025
Closed-form norm scaling with data for overparameterized linear regression and diagonal linear networks under bias
Shuofeng Zhang, Ard Louis
For overparameterized linear regression with isotropic Gaussian design and minimum- interpolator , we give a unified, high-probability characterization for the s…
cs.LG2021
Why flatness does and does not correlate with generalization for deep neural networks
Shuofeng Zhang, Isaac Reid, Guillermo Valle Pérez +1
The intuition that local flatness of the loss landscape is correlated with better generalization for deep neural networks (DNNs) has been explored for decades, spawning many differ…