Zeroth-Order Hard-Thresholding: Gradient Error vs. Expansivity

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: de Vazelhes, William, Zhang, Hualin, Wu, Huimin, Yuan, Xiao-Tong, Gu, Bin
Natura: Preprint
Pubblicazione: 2022
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910369377681408
author de Vazelhes, William
Zhang, Hualin
Wu, Huimin
Yuan, Xiao-Tong
Gu, Bin
author_facet de Vazelhes, William
Zhang, Hualin
Wu, Huimin
Yuan, Xiao-Tong
Gu, Bin
contents $\ell_0$ constrained optimization is prevalent in machine learning, particularly for high-dimensional problems, because it is a fundamental approach to achieve sparse learning. Hard-thresholding gradient descent is a dominant technique to solve this problem. However, first-order gradients of the objective function may be either unavailable or expensive to calculate in a lot of real-world problems, where zeroth-order (ZO) gradients could be a good surrogate. Unfortunately, whether ZO gradients can work with the hard-thresholding operator is still an unsolved problem. To solve this puzzle, in this paper, we focus on the $\ell_0$ constrained black-box stochastic optimization problems, and propose a new stochastic zeroth-order gradient hard-thresholding (SZOHT) algorithm with a general ZO gradient estimator powered by a novel random support sampling. We provide the convergence analysis of SZOHT under standard assumptions. Importantly, we reveal a conflict between the deviation of ZO estimators and the expansivity of the hard-thresholding operator, and provide a theoretical minimal value of the number of random directions in ZO gradients. In addition, we find that the query complexity of SZOHT is independent or weakly dependent on the dimensionality under different settings. Finally, we illustrate the utility of our method on a portfolio optimization problem as well as black-box adversarial attacks.
format Preprint
id arxiv_https___arxiv_org_abs_2210_05279
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Zeroth-Order Hard-Thresholding: Gradient Error vs. Expansivity
de Vazelhes, William
Zhang, Hualin
Wu, Huimin
Yuan, Xiao-Tong
Gu, Bin
Machine Learning
Optimization and Control
$\ell_0$ constrained optimization is prevalent in machine learning, particularly for high-dimensional problems, because it is a fundamental approach to achieve sparse learning. Hard-thresholding gradient descent is a dominant technique to solve this problem. However, first-order gradients of the objective function may be either unavailable or expensive to calculate in a lot of real-world problems, where zeroth-order (ZO) gradients could be a good surrogate. Unfortunately, whether ZO gradients can work with the hard-thresholding operator is still an unsolved problem. To solve this puzzle, in this paper, we focus on the $\ell_0$ constrained black-box stochastic optimization problems, and propose a new stochastic zeroth-order gradient hard-thresholding (SZOHT) algorithm with a general ZO gradient estimator powered by a novel random support sampling. We provide the convergence analysis of SZOHT under standard assumptions. Importantly, we reveal a conflict between the deviation of ZO estimators and the expansivity of the hard-thresholding operator, and provide a theoretical minimal value of the number of random directions in ZO gradients. In addition, we find that the query complexity of SZOHT is independent or weakly dependent on the dimensionality under different settings. Finally, we illustrate the utility of our method on a portfolio optimization problem as well as black-box adversarial attacks.
title Zeroth-Order Hard-Thresholding: Gradient Error vs. Expansivity
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2210.05279