{"id":"arxiv/2610.30617","url":"https://arxiv.org/abs/2610.30617","tags":["math.OC"],"score":0,"title":"Retraction-Free Nonsmooth Optimization on the Stiefel Manifold: Subgradient and Proximal Landing Methods","domain":"math.OC","source":"arxiv","authors":["Zilong Cui","Tongxin Hu","Bin Gao","Ran Gu"],"reviews":0,"subject":"math.OC","summary":"We develop three retraction-free Landing methods for nonsmooth composite optimization over the Stiefel manifold, combining an (approximately) tangential direction with an infeasibility-reducing normal direction. The subgradient Landing method, with normal stepsizes parameterized by $α\\in[0,1]$, has best-iterate complexity bounds $\\mathcal{O}(ε^{-4})$ for surrogate stationarity and $\\mathcal{O}(ε^{-2/(2-α)})$ for feasibility. For $α=0$, it converges locally $R$-linearly under local weak sharpness and mild assumptions. The proximal gradient Landing method computes a tangential direction by…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-22","reactions":{}}