arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.09209math.OCcs.AI
收藏

选择后再求导:解决具有流形下层解集的双层优化问题

Select-then-differentiate: Solving Bilevel Optimization with Manifold Lower-level Solution Sets

EPFL School of Management of Technology(EPFL管理学院);ETH Department of Computer Science(ETH计算机科学系)

Saeed Masiha, Zebang Shen, Negar Kiyavash, Niao He

AI总结:

本文研究了下层问题具有非孤立流形极小解时的乐观双层优化,提出HG-MS方法结合显式乐观选择和高效的伪逆超梯度计算,证明其在流形内在维度内收敛,并在实际应用中取得最佳成绩。

摘要:

我们研究了当下层问题具有非孤立流形极小解时的乐观双层优化。在该设定下,超目标可能非可导,因为上层准则必须在多个下层解中选择。在局部Polyak--Łojasiewicz (PŁ)条件下,我们证明不同可导性不需要下层解集为单例:乐观选择的唯一性足以。这导致了基于伪逆的显式超梯度公式,扩展了经典单例极小解结果。我们进一步表征了超目标的正则性:所选极小解沿解流形的非退化性导致局部光滑性,而唯一性失败可产生多个非可导点,非退化性失败可破坏所有正则超梯度的正则性。受此理论启发,我们提出HG-MS,一种结合显式乐观选择与高效伪逆超梯度计算的选然后求导方法。尽管乐观选择在下层解流形上非凸,我们证明HG-MS收敛到乐观目标的 stationary 点,其复杂性由解流形的内在维度而非其嵌入维度决定。实证上,我们测试了HG-MS的实用变体用于匹配预算的LLM源权重重,该变体保持选然后求导原则,并在测试的backbones上获得最佳GSM8K/MATH分数,以及竞争或最佳的MT-Bench指令遵循结果。

英文摘要:

We study optimistic bilevel optimization when the lower-level problem has a non-isolated manifold of minimizers. In this setting, the hyper-objective may be non-differentiable because the upper-level criterion must choose among multiple lower-level solutions. Under a local Polyak--Łojasiewicz (PŁ) condition, we show that differentiability does not require the lower-level solution set to be a singleton: uniqueness of the optimistic selection is sufficient. This yields an explicit pseudoinverse-based hyper-gradient formula extending the classical singleton-minimizer result. We further characterize the regularity of the hyper-objective: non-degeneracy of the selected minimizer along the solution manifold yields local smoothness, while failure of uniqueness can create many non-differentiable points and failure of non-degeneracy can destroy all positive Hölder regularity of the hyper-gradient. Motivated by this theory, we propose HG-MS, a select-then-differentiate method combining explicit optimistic selection with efficient pseudoinverse-based hyper-gradient computation. Despite the nonconvex nature of optimistic selection over the lower-level solution manifold, we show that HG-MS converges to a stationary point of the optimistic objective with complexity governed by the intrinsic dimension of the solution manifold rather than its ambient dimension. Empirically, we test a practical variant of HG-MS for matched-budget LLM source reweighting. This variant preserves the select-then-differentiate principle and obtains the best GSM8K/MATH scores across the tested backbones, along with competitive or best MT-Bench instruction-following results.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑