{"id":"arxiv/2609.17229","url":"https://arxiv.org/abs/2609.17229","tags":["math.NA"],"score":0,"title":"A provably convergent MM-GKS variant for large-scale inverse problems","domain":"math.NA","source":"arxiv","authors":["Mirjeta Pasha","Eric de Sturler","Misha Kilmer"],"reviews":0,"subject":"math.NA","summary":"For high-quality images with sharp edges, a popular choice for edge-preserving regularization is using a general(ized) $\\ell_q$-norm of the gradient of the image. This can be implemented efficiently using the $\\ell_2$-norm and a sequence of weighted gradients, with weights derived from the current solution estimate. We can solve the resulting sequence of regularized least squares problems using hybrid Krylov subspace methods, which efficiently compute the regularization parameter using the problem projected on the Krylov subspace. However, each update of the regularization operator requires…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-15","reactions":{}}