{"id":"arxiv/2609.30885","url":"https://arxiv.org/abs/2609.30885","tags":["math.OC","cs.LG"],"score":0,"title":"Retraction-Based Gradient Projection Algorithms on Manifolds","domain":"math.OC","source":"arxiv","authors":["Conglong Xu","Hao Wu"],"reviews":0,"subject":"math.OC","summary":"We introduce a framework for retraction-based convex optimization on Riemannian manifolds, which includes a notion of retraction-specific convex sets and retraction-based gradient projection algorithms. The standard theory of gradient projection algorithms generalizes easily to this framework. Within this framework, we establish convergence results for retraction-based gradient projection algorithms with various stepsize rules. As an application, we use our framework to study the weighted low-rank approximation. We also provide numerical validation of our convergence results on the image…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-25","reactions":{}}