Solving bilevel optimization via sequential minimax optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lu, Zhaosong, Mei, Sanyou
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915610216103936
author Lu, Zhaosong
Mei, Sanyou
author_facet Lu, Zhaosong
Mei, Sanyou
contents In this paper we propose a sequential minimax optimization (SMO) method for solving a class of constrained bilevel optimization problems in which the lower-level part is a possibly nonsmooth convex optimization problem, while the upper-level part is a possibly nonconvex optimization problem. Specifically, SMO applies a first-order method to solve a sequence of minimax subproblems, which are obtained by employing a hybrid of modified augmented Lagrangian and penalty schemes on the bilevel optimization problems. Under suitable assumptions, we establish an operation complexity of $O(\varepsilon^{-7}\log\varepsilon^{-1})$ and $O(\varepsilon^{-6}\log\varepsilon^{-1})$, measured in terms of fundamental operations, for SMO in finding an $\varepsilon$-KKT solution of the bilevel optimization problems with merely convex and strongly convex lower-level objective functions, respectively. The latter result improves the previous best-known operation complexity by a factor of $\varepsilon^{-1}$. Preliminary numerical results demonstrate significantly superior computational performance compared to the recently developed first-order penalty method.
format Preprint
id arxiv_https___arxiv_org_abs_2511_07398
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Solving bilevel optimization via sequential minimax optimization
Lu, Zhaosong
Mei, Sanyou
Optimization and Control
Machine Learning
Numerical Analysis
90C26, 90C30, 90C47, 90C99, 65K05
In this paper we propose a sequential minimax optimization (SMO) method for solving a class of constrained bilevel optimization problems in which the lower-level part is a possibly nonsmooth convex optimization problem, while the upper-level part is a possibly nonconvex optimization problem. Specifically, SMO applies a first-order method to solve a sequence of minimax subproblems, which are obtained by employing a hybrid of modified augmented Lagrangian and penalty schemes on the bilevel optimization problems. Under suitable assumptions, we establish an operation complexity of $O(\varepsilon^{-7}\log\varepsilon^{-1})$ and $O(\varepsilon^{-6}\log\varepsilon^{-1})$, measured in terms of fundamental operations, for SMO in finding an $\varepsilon$-KKT solution of the bilevel optimization problems with merely convex and strongly convex lower-level objective functions, respectively. The latter result improves the previous best-known operation complexity by a factor of $\varepsilon^{-1}$. Preliminary numerical results demonstrate significantly superior computational performance compared to the recently developed first-order penalty method.
title Solving bilevel optimization via sequential minimax optimization
topic Optimization and Control
Machine Learning
Numerical Analysis
90C26, 90C30, 90C47, 90C99, 65K05
url https://arxiv.org/abs/2511.07398