Current browse context:
math.OC
Change to browse by:
References & Citations
Mathematics > Optimization and Control
Title: Accelerated Block Coordinate Proximal Gradients with Applications in High Dimensional Statistics
(Submitted on 15 Oct 2017 (v1), revised 17 Oct 2017 (this version, v2), latest version 3 Dec 2017 (v7))
Abstract: Nonconvex optimization problems arise in different research fields and arouse lots of attention in signal processing, statistics and machine learning. In this work, we explore the accelerated proximal gradient method and some of its variants which have been shown to converge under nonconvex context recently. We show that a novel variant proposed here, which exploits adaptive momentum and block coordinate update with specific update rules, further improves the performance of a broad class of nonconvex problems. In applications to sparse linear regression with regularizations like Lasso, grouped Lasso, capped $\ell_1$ and SCAP, the proposed scheme enjoys provable local linear convergence, with experimental justification.
Submission history
From: Tsz Kit Lau [view email][v1] Sun, 15 Oct 2017 14:07:32 GMT (231kb,D)
[v2] Tue, 17 Oct 2017 11:21:32 GMT (231kb,D)
[v3] Fri, 27 Oct 2017 06:31:38 GMT (225kb,D)
[v4] Mon, 30 Oct 2017 15:16:24 GMT (225kb,D)
[v5] Tue, 31 Oct 2017 11:40:24 GMT (225kb,D)
[v6] Sat, 18 Nov 2017 09:26:25 GMT (192kb,D)
[v7] Sun, 3 Dec 2017 11:21:03 GMT (191kb,D)
Link back to: arXiv, form interface, contact.