English

Stochastic optimization over expectation-formulated generalized Stiefel manifold

Optimization and Control 2024-12-31 v1

Abstract

In this paper, we consider a class of stochastic optimization problems over the expectation-formulated generalized Stiefel manifold (SOEGS), where the objective function ff is continuously differentiable. We propose a novel constraint dissolving penalty function with a customized penalty term (CDFDP), which maintains the same order of differentiability as ff. Our theoretical analysis establishes the global equivalence between CDFCP and SOEGS in the sense that they share the same first-order and second-order stationary points under mild conditions. These results on equivalence enable the direct implementation of various stochastic optimization approaches to solve SOEGS. In particular, we develop a stochastic gradient algorithm and its accelerated variant by incorporating an adaptive step size strategy. Furthermore, we prove their O(ε4)\mathcal{O}(\varepsilon^{-4}) sample complexity for finding an ε\varepsilon-stationary point of CDFCP. Comprehensive numerical experiments show the efficiency and robustness of our proposed algorithms.

Keywords

Cite

@article{arxiv.2412.20008,
  title  = {Stochastic optimization over expectation-formulated generalized Stiefel manifold},
  author = {Linshuo Jiang and Nachuan Xiao and Xin Liu},
  journal= {arXiv preprint arXiv:2412.20008},
  year   = {2024}
}

Comments

34 pages

R2 v1 2026-06-28T20:50:26.404Z