[2402.16712] l1-norm regularized l1-norm best-fit strains

0
21


Obtain a PDF of the paper titled l1-norm regularized l1-norm best-fit strains, by Xiao Ling and 1 different authors

Obtain PDF

Summary:On this work, we suggest an optimization framework for estimating a sparse strong one-dimensional subspace. Our goal is to attenuate each the illustration error and the penalty, when it comes to the l1-norm criterion. Provided that the issue is NP-hard, we introduce a linear relaxation-based method. Moreover, we current a novel becoming process, using easy ratios and sorting methods. The proposed algorithm demonstrates a worst-case time complexity of $O(n^2 m log n)$ and, in sure situations, achieves world optimality for the sparse strong subspace, thereby exhibiting polynomial time effectivity. In comparison with extant methodologies, the proposed algorithm finds the subspace with the bottom discordance, providing a smoother trade-off between sparsity and match. Its structure affords scalability, evidenced by a 16-fold enchancment in computational speeds for matrices of 2000×2000 over CPU model. Moreover, this methodology is distinguished by a number of benefits, together with its independence from initialization and deterministic and replicable procedures. Moreover, this methodology is distinguished by a number of benefits, together with its independence from initialization and deterministic and replicable procedures. The true-world instance demonstrates the effectiveness of algorithm in attaining significant sparsity, underscoring its exact and helpful utility throughout numerous domains.

Submission historical past

From: Xiao Ling [view email]
[v1]
Mon, 26 Feb 2024 16:30:58 UTC (81 KB)
[v2]
Wed, 6 Mar 2024 17:16:38 UTC (81 KB)



Supply hyperlink

LEAVE A REPLY

Please enter your comment!
Please enter your name here