Near-Optimal Nonconvex-Strongly-Convex Bilevel Optimization with Fully First-Order Oracles
Authors
Research Topics
Paper Information
-
Journal:
Journal of Machine Learning Research -
Added to Tracker:
Jul 15, 2025
Abstract
In this work, we consider bilevel optimization when the lower-level problem is strongly convex. Recent works show that with a Hessian-vector product (HVP) oracle, one can provably find an $\epsilon$-stationary point within ${O}(\epsilon^{-2})$ oracle calls. However, the HVP oracle may be inaccessible or expensive in practice. Kwon et al. (ICML 2023) addressed this issue by proposing a first-order method that can achieve the same goal at a slower rate of $\tilde{O}(\epsilon^{-3})$. In this paper, we incorporate a two-time-scale update to improve their method to achieve the near-optimal $\tilde{O}(\epsilon^{-2})$ first-order oracle complexity. Our analysis is highly extensible. In the stochastic setting, our algorithm can achieve the stochastic first-order oracle complexity of $\tilde {O}(\epsilon^{-4})$ and $\tilde {O}(\epsilon^{-6})$ when the stochastic noises are only in the upper-level objective and in both level objectives, respectively. When the objectives have higher-order smoothness conditions, our deterministic method can escape saddle points by injecting noise, and can be accelerated to achieve a faster rate of $\tilde {O}(\epsilon^{-1.75})$ using Nesterov's momentum.
Author Details
Lesi Chen
AuthorYaohua Ma
AuthorJingzhao Zhang
AuthorResearch Topics & Keywords
Computational Statistics
Research AreaCitation Information
APA Format
Lesi Chen
,
Yaohua Ma
&
Jingzhao Zhang
.
Near-Optimal Nonconvex-Strongly-Convex Bilevel Optimization with Fully First-Order Oracles.
Journal of Machine Learning Research
.
BibTeX Format
@article{JMLR:v26:23-1104,
author = {Lesi Chen and Yaohua Ma and Jingzhao Zhang},
title = {Near-Optimal Nonconvex-Strongly-Convex Bilevel Optimization with Fully First-Order Oracles},
journal = {Journal of Machine Learning Research},
year = {2025},
volume = {26},
number = {109},
pages = {1--56},
url = {http://jmlr.org/papers/v26/23-1104.html}
}