The Full Spectrum of Deepnet Hessians at Scale: Dynamics with SGD Training and Sample Size
Machine Learning
2019-06-04 v2 Machine Learning
Abstract
We apply state-of-the-art tools in modern high-dimensional numerical linear algebra to approximate efficiently the spectrum of the Hessian of modern deepnets, with tens of millions of parameters, trained on real data. Our results corroborate previous findings, based on small-scale networks, that the Hessian exhibits "spiked" behavior, with several outliers isolated from a continuous bulk. We decompose the Hessian into different components and study the dynamics with training and sample size of each term individually.
Keywords
Cite
@article{arxiv.1811.07062,
title = {The Full Spectrum of Deepnet Hessians at Scale: Dynamics with SGD Training and Sample Size},
author = {Vardan Papyan},
journal= {arXiv preprint arXiv:1811.07062},
year = {2019}
}