The Randomized Block Coordinate Descent Method in the Hölder Smooth Setting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Maia, Leandro Farias, Gutman, David Huckleberry
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929274064207872
author Maia, Leandro Farias
Gutman, David Huckleberry
author_facet Maia, Leandro Farias
Gutman, David Huckleberry
contents This work provides the first convergence analysis for the Randomized Block Coordinate Descent method for minimizing a function that is both Hölder smooth and block Hölder smooth. Our analysis applies to objective functions that are non-convex, convex, and strongly convex. For non-convex functions, we show that the expected gradient norm reduces at an $O\left(k^{\fracγ{1+γ}}\right)$ rate, where $k$ is the iteration count and $γ$ is the Hölder exponent. For convex functions, we show that the expected suboptimality gap reduces at the rate $O\left(k^{-γ}\right)$. In the strongly convex setting, we show this rate for the expected suboptimality gap improves to $O\left(k^{-\frac{2γ}{1-γ}}\right)$ when $γ>1$ and to a linear rate when $γ=1$. Notably, these new convergence rates coincide with those furnished in the existing literature for the Lipschitz smooth setting.
format Preprint
id arxiv_https___arxiv_org_abs_2403_08080
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Randomized Block Coordinate Descent Method in the Hölder Smooth Setting
Maia, Leandro Farias
Gutman, David Huckleberry
Optimization and Control
This work provides the first convergence analysis for the Randomized Block Coordinate Descent method for minimizing a function that is both Hölder smooth and block Hölder smooth. Our analysis applies to objective functions that are non-convex, convex, and strongly convex. For non-convex functions, we show that the expected gradient norm reduces at an $O\left(k^{\fracγ{1+γ}}\right)$ rate, where $k$ is the iteration count and $γ$ is the Hölder exponent. For convex functions, we show that the expected suboptimality gap reduces at the rate $O\left(k^{-γ}\right)$. In the strongly convex setting, we show this rate for the expected suboptimality gap improves to $O\left(k^{-\frac{2γ}{1-γ}}\right)$ when $γ>1$ and to a linear rate when $γ=1$. Notably, these new convergence rates coincide with those furnished in the existing literature for the Lipschitz smooth setting.
title The Randomized Block Coordinate Descent Method in the Hölder Smooth Setting
topic Optimization and Control
url https://arxiv.org/abs/2403.08080