Gradient Descent Converges to Minimizers
Machine Learning
2016-03-07 v2 Machine Learning
Optimization and Control
Abstract
We show that gradient descent converges to a local minimizer, almost surely with random initialization. This is proved by applying the Stable Manifold Theorem from dynamical systems theory.
Cite
@article{arxiv.1602.04915,
title = {Gradient Descent Converges to Minimizers},
author = {Jason D. Lee and Max Simchowitz and Michael I. Jordan and Benjamin Recht},
journal= {arXiv preprint arXiv:1602.04915},
year = {2016}
}
Comments
Submitted to COLT 2016