Saved in:
Bibliographic Details
Main Authors: Sivan, Hadar, Gabel, Moshe, Schuster, Assaf
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2302.08484
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914706476761088
author Sivan, Hadar
Gabel, Moshe
Schuster, Assaf
author_facet Sivan, Hadar
Gabel, Moshe
Schuster, Assaf
contents Popular machine learning approaches forgo second-order information due to the difficulty of computing curvature in high dimensions. We present FOSI, a novel meta-algorithm that improves the performance of any base first-order optimizer by efficiently incorporating second-order information during the optimization process. In each iteration, FOSI implicitly splits the function into two quadratic functions defined on orthogonal subspaces, then uses a second-order method to minimize the first, and the base optimizer to minimize the other. We formally analyze FOSI's convergence and the conditions under which it improves a base optimizer. Our empirical evaluation demonstrates that FOSI improves the convergence rate and optimization time of first-order methods such as Heavy-Ball and Adam, and outperforms second-order methods (K-FAC and L-BFGS).
format Preprint
id arxiv_https___arxiv_org_abs_2302_08484
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle FOSI: Hybrid First and Second Order Optimization
Sivan, Hadar
Gabel, Moshe
Schuster, Assaf
Machine Learning
Popular machine learning approaches forgo second-order information due to the difficulty of computing curvature in high dimensions. We present FOSI, a novel meta-algorithm that improves the performance of any base first-order optimizer by efficiently incorporating second-order information during the optimization process. In each iteration, FOSI implicitly splits the function into two quadratic functions defined on orthogonal subspaces, then uses a second-order method to minimize the first, and the base optimizer to minimize the other. We formally analyze FOSI's convergence and the conditions under which it improves a base optimizer. Our empirical evaluation demonstrates that FOSI improves the convergence rate and optimization time of first-order methods such as Heavy-Ball and Adam, and outperforms second-order methods (K-FAC and L-BFGS).
title FOSI: Hybrid First and Second Order Optimization
topic Machine Learning
url https://arxiv.org/abs/2302.08484