Saved in:
Bibliographic Details
Main Authors: Thomsen, Daniel Berg, Doikov, Nikita
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2404.17543
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909466050428928
author Thomsen, Daniel Berg
Doikov, Nikita
author_facet Thomsen, Daniel Berg
Doikov, Nikita
contents In this work, we study the iteration complexity of gradient methods for minimizing convex quadratic functions regularized by powers of Euclidean norms. We show that, due to the uniform convexity of the objective, gradient methods have improved convergence rates. Thus, for the basic gradient descent with a novel step size, we prove a convergence rate of $O(N^{-p/(p - 2)})$ for the functional residual, where $N$ is the iteration number and $p > 2$ is the power of the regularization term. We also show that this rate is tight by establishing a corresponding lower bound for one-step first-order methods. Then, for the general class of all multi-step methods, we establish that the rate of $O(N^{-2p/(p-2)})$ is optimal, providing a sharp analysis of the minimization of uniformly convex regularized quadratic functions. This rate is achieved by the fast gradient method. A special case of our problem class is $p=3$, which is the minimization of cubically regularized convex quadratic functions. It naturally appears as a subproblem at each iteration of the cubic Newton method. Therefore, our theory shows that the rate of $O(N^{-6})$ is optimal in this case. We also establish new lower bounds on minimizing the gradient norm within our framework.
format Preprint
id arxiv_https___arxiv_org_abs_2404_17543
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Complexity of Minimizing Regularized Convex Quadratic Functions
Thomsen, Daniel Berg
Doikov, Nikita
Optimization and Control
In this work, we study the iteration complexity of gradient methods for minimizing convex quadratic functions regularized by powers of Euclidean norms. We show that, due to the uniform convexity of the objective, gradient methods have improved convergence rates. Thus, for the basic gradient descent with a novel step size, we prove a convergence rate of $O(N^{-p/(p - 2)})$ for the functional residual, where $N$ is the iteration number and $p > 2$ is the power of the regularization term. We also show that this rate is tight by establishing a corresponding lower bound for one-step first-order methods. Then, for the general class of all multi-step methods, we establish that the rate of $O(N^{-2p/(p-2)})$ is optimal, providing a sharp analysis of the minimization of uniformly convex regularized quadratic functions. This rate is achieved by the fast gradient method. A special case of our problem class is $p=3$, which is the minimization of cubically regularized convex quadratic functions. It naturally appears as a subproblem at each iteration of the cubic Newton method. Therefore, our theory shows that the rate of $O(N^{-6})$ is optimal in this case. We also establish new lower bounds on minimizing the gradient norm within our framework.
title Complexity of Minimizing Regularized Convex Quadratic Functions
topic Optimization and Control
url https://arxiv.org/abs/2404.17543