A Retraction-free Method for Nonsmooth Minimax Optimization over a Compact Manifold
Abstract
We study the minimax problem , where is a compact submanifold, is continuously differentiable in , is a closed, weakly-convex (possibly non-smooth) function and we assume that the regularized coupling function is either -PL for some or concave () for any fixed in the vicinity of . To address the nonconvexity due to the manifold constraint, we use an exact penalty for the constraint , and enforcing a convex constraint for some , onto which projections can be computed efficiently. Building upon this new formulation for the manifold minimax problem in question, a single-loop smoothed manifold gradient descent-ascent (sm-MGDA) algorithm is proposed. Theoretically, any limit point of sm-MGDA sequence is a stationary point of the manifold minimax problem and sm-MGDA can generate an -stationary point of the original problem with and complexity for and scenarios, respectively. Moreover, for the setting, through adopting Tikhonov regularization of the dual, one can improve the complexity to at the expense of asymptotic stationarity. The key component, common in the analysis of all cases, is to connect -stationary points between the penalized problem and the original problem by showing that the constraint becomes inactive and the penalty term tends to along any convergent subsequence. To our knowledge, sm-MGDA is the first retraction-free algorithm for minimax problems over compact submanifolds, and this is a very desirable algorithmic property since through avoiding retractions, one can get away with matrix orthogonalization subroutines required for computing retractions to manifolds arising in practice, which are not GPU friendly.
Cite
@article{arxiv.2510.22065,
title = {A Retraction-free Method for Nonsmooth Minimax Optimization over a Compact Manifold},
author = {Necdet Serhat Aybat and Jiang Hu and Zhanwang Deng},
journal= {arXiv preprint arXiv:2510.22065},
year = {2025}
}
Comments
New numerical results are added