Score: 2

A Single-Loop First-Order Algorithm for Linearly Constrained Bilevel Optimization

Published: October 28, 2025 | arXiv ID: 2510.24710v1

By: Wei Shen , Jiawei Zhang , Minhui Huang and more

BigTech Affiliations: Meta

Potential Business Impact:

Solves tricky math problems faster than before.

Business Areas:
A/B Testing Data and Analytics

We study bilevel optimization problems where the lower-level problems are strongly convex and have coupled linear constraints. To overcome the potential non-smoothness of the hyper-objective and the computational challenges associated with the Hessian matrix, we utilize penalty and augmented Lagrangian methods to reformulate the original problem as a single-level one. Especially, we establish a strong theoretical connection between the reformulated function and the original hyper-objective by characterizing the closeness of their values and derivatives. Based on this reformulation, we propose a single-loop, first-order algorithm for linearly constrained bilevel optimization (SFLCB). We provide rigorous analyses of its non-asymptotic convergence rates, showing an improvement over prior double-loop algorithms -- form $O(\epsilon^{-3}\log(\epsilon^{-1}))$ to $O(\epsilon^{-3})$. The experiments corroborate our theoretical findings and demonstrate the practical efficiency of the proposed SFLCB algorithm. Simulation code is provided at https://github.com/ShenGroup/SFLCB.

Country of Origin
πŸ‡ΊπŸ‡Έ United States

Repos / Data Links

Page Count
40 pages

Category
Mathematics:
Optimization and Control