Saved in:
Bibliographic Details
Main Authors: Liu, Yu, Peng, Weibin, Wang, Tianyu, Yu, Jiajia
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.22357
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912531851771904
author Liu, Yu
Peng, Weibin
Wang, Tianyu
Yu, Jiajia
author_facet Liu, Yu
Peng, Weibin
Wang, Tianyu
Yu, Jiajia
contents This paper studies stochastic minimization of a finite-sum loss $ F (\mathbf{x}) = \frac{1}{N} \sum_{ξ=1}^N f(\mathbf{x};ξ) $. In many real-world scenarios, the Hessian matrix of such objectives exhibits a low-rank structure on a batch of data. At the same time, zeroth-order optimization has gained prominence in important applications such as fine-tuning large language models. Drawing on these observations, we propose a novel stochastic zeroth-order cubic Newton method that leverages the low-rank Hessian structure via a matrix recovery-based estimation technique. Our method circumvents restrictive incoherence assumptions, enabling accurate Hessian approximation through finite-difference queries. Theoretically, we establish that for most real-world problems in $\mathbb{R}^n$, $\mathcal{O}\left(\frac{n}{η^{\frac{7}{2}}}\right)+\widetilde{\mathcal{O}}\left(\frac{n^2 }{η^{\frac{5}{2}}}\right)$ function evaluations suffice to attain a second-order $η$-stationary point with high probability. This represents a significant improvement in dimensional dependence over existing methods. This improvement is mostly due to a new Hessian estimator that achieves superior sample complexity; This new Hessian estimation method might be of separate interest. Numerical experiments on matrix recovery and machine learning tasks validate the efficacy and scalability of our approach.
format Preprint
id arxiv_https___arxiv_org_abs_2410_22357
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Zeroth-order Stochastic Cubic Newton Method Revisited
Liu, Yu
Peng, Weibin
Wang, Tianyu
Yu, Jiajia
Optimization and Control
This paper studies stochastic minimization of a finite-sum loss $ F (\mathbf{x}) = \frac{1}{N} \sum_{ξ=1}^N f(\mathbf{x};ξ) $. In many real-world scenarios, the Hessian matrix of such objectives exhibits a low-rank structure on a batch of data. At the same time, zeroth-order optimization has gained prominence in important applications such as fine-tuning large language models. Drawing on these observations, we propose a novel stochastic zeroth-order cubic Newton method that leverages the low-rank Hessian structure via a matrix recovery-based estimation technique. Our method circumvents restrictive incoherence assumptions, enabling accurate Hessian approximation through finite-difference queries. Theoretically, we establish that for most real-world problems in $\mathbb{R}^n$, $\mathcal{O}\left(\frac{n}{η^{\frac{7}{2}}}\right)+\widetilde{\mathcal{O}}\left(\frac{n^2 }{η^{\frac{5}{2}}}\right)$ function evaluations suffice to attain a second-order $η$-stationary point with high probability. This represents a significant improvement in dimensional dependence over existing methods. This improvement is mostly due to a new Hessian estimator that achieves superior sample complexity; This new Hessian estimation method might be of separate interest. Numerical experiments on matrix recovery and machine learning tasks validate the efficacy and scalability of our approach.
title Zeroth-order Stochastic Cubic Newton Method Revisited
topic Optimization and Control
url https://arxiv.org/abs/2410.22357