Generalization in Supervised Learning Through Riemannian Contraction
Machine Learning
2022-01-27 v2 Dynamical Systems
Abstract
We prove that Riemannian contraction in a supervised learning setting implies generalization. Specifically, we show that if an optimizer is contracting in some Riemannian metric with rate , it is uniformly algorithmically stable with rate , where is the number of labelled examples in the training set. The results hold for stochastic and deterministic optimization, in both continuous and discrete-time, for convex and non-convex loss surfaces. The associated generalization bounds reduce to well-known results in the particular case of gradient descent over convex or strongly convex loss surfaces. They can be shown to be optimal in certain linear settings, such as kernel ridge regression under gradient flow.
Cite
@article{arxiv.2201.06656,
title = {Generalization in Supervised Learning Through Riemannian Contraction},
author = {Leo Kozachkov and Patrick M. Wensing and Jean-Jacques Slotine},
journal= {arXiv preprint arXiv:2201.06656},
year = {2022}
}
Comments
22 pages, 5, figures