Achieving ${O}(ε^{-1.5})$ Complexity in Hessian/Jacobian-free Stochastic Bilevel Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Yifan, Xiao, Peiyao, Ji, Kaiyi
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908303126167552
author Yang, Yifan
Xiao, Peiyao
Ji, Kaiyi
author_facet Yang, Yifan
Xiao, Peiyao
Ji, Kaiyi
contents In this paper, we revisit the bilevel optimization problem, in which the upper-level objective function is generally nonconvex and the lower-level objective function is strongly convex. Although this type of problem has been studied extensively, it still remains an open question how to achieve an ${O}(ε^{-1.5})$ sample complexity in Hessian/Jacobian-free stochastic bilevel optimization without any second-order derivative computation. To fill this gap, we propose a novel Hessian/Jacobian-free bilevel optimizer named FdeHBO, which features a simple fully single-loop structure, a projection-aided finite-difference Hessian/Jacobian-vector approximation, and momentum-based updates. Theoretically, we show that FdeHBO requires ${O}(ε^{-1.5})$ iterations (each using ${O}(1)$ samples and only first-order gradient information) to find an $ε$-accurate stationary point. As far as we know, this is the first Hessian/Jacobian-free method with an ${O}(ε^{-1.5})$ sample complexity for nonconvex-strongly-convex stochastic bilevel optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2312_03807
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Achieving ${O}(ε^{-1.5})$ Complexity in Hessian/Jacobian-free Stochastic Bilevel Optimization
Yang, Yifan
Xiao, Peiyao
Ji, Kaiyi
Optimization and Control
Machine Learning
In this paper, we revisit the bilevel optimization problem, in which the upper-level objective function is generally nonconvex and the lower-level objective function is strongly convex. Although this type of problem has been studied extensively, it still remains an open question how to achieve an ${O}(ε^{-1.5})$ sample complexity in Hessian/Jacobian-free stochastic bilevel optimization without any second-order derivative computation. To fill this gap, we propose a novel Hessian/Jacobian-free bilevel optimizer named FdeHBO, which features a simple fully single-loop structure, a projection-aided finite-difference Hessian/Jacobian-vector approximation, and momentum-based updates. Theoretically, we show that FdeHBO requires ${O}(ε^{-1.5})$ iterations (each using ${O}(1)$ samples and only first-order gradient information) to find an $ε$-accurate stationary point. As far as we know, this is the first Hessian/Jacobian-free method with an ${O}(ε^{-1.5})$ sample complexity for nonconvex-strongly-convex stochastic bilevel optimization.
title Achieving ${O}(ε^{-1.5})$ Complexity in Hessian/Jacobian-free Stochastic Bilevel Optimization
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2312.03807