Id: arxiv/2611.22940 Url: https://arxiv.org/abs/2611.22940 Tags: math.OC Score: 0 Title: KL-Guided Support Selection and Subspace Cubic Newton Methods for Sparse Optimization Domain: math.OC Source: arxiv Authors: Zihao Xia, Min Tao, Guojian Yin Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-11-17 Page: https://delphi.science/r/arxiv/2611.22940 We study sparse composite optimization with a smooth loss and a possibly nonconvex, nonsmooth regularizer. The objective is locally smooth on each fixed support. Our method combines full-space proximal-gradient steps with inexact cubic-regularized Newton steps on a subspace selected by thresholding. If the iterates are bounded and the objective satisfies the Kurdyka--Łojasiewicz (KL) property, the entire sequence converges to a first-order stationary point. A threshold based on the KL exponent ensures that the selected coordinate set coincides with the support of the limit after finitely…