Escaping Saddles with Stochastic Gradients
Abstract
We analyze the variance of stochastic gradients along negative curvature directions in certain non-convex machine learning models and show that stochastic gradients exhibit a strong component along these directions. Furthermore, we show that - contrary to the case of isotropic noise - this variance is proportional to the magnitude of the corresponding eigenvalues and not decreasing in the dimensionality. Based upon this observation we propose a new assumption under which we show that the injection of explicit, isotropic noise usually applied to make gradient descent escape saddle points can successfully be replaced by a simple SGD step. Additionally - and under the same condition - we derive the first convergence rate for plain SGD to a second-order stationary point in a number of iterations that is independent of the problem dimension.
Cite
@article{arxiv.1803.05999,
title = {Escaping Saddles with Stochastic Gradients},
author = {Hadi Daneshmand and Jonas Kohler and Aurelien Lucchi and Thomas Hofmann},
journal= {arXiv preprint arXiv:1803.05999},
year = {2018}
}