Faster than Fast-LTS: Robust Regression and Outlier Detection with DC Programming

DOI10.48550/arXiv.2606.28974MaRDI QIDQ7270460arXivFDO

Authors Marah-Lisanne Thormann, Phan Tu Vuong, Alain B. Zemkoho, Tri-Dung Nguyen

Publication date 27 June 2026

Published in arXiv

Tldr We reformulate LTS as a concave minimization problem subject to a capped simplex constraint and propose the successive Boosted Difference of Convex Functions Algorithm (sBDCA) with preconditioning for solving it efficiently. We prove sBDCA's linear convergence rate via Łojasiewicz property, outperform Fast-LTS by up to 3.25 times in terms of computation time and achieving up to 90% lower objective function values on various synthetic and real-world datasets.

Full work available at URL https://doi.org/10.48550/arXiv.2606.28974

arXiv classification

math.NAmath.OCmath.ST

This page was built for publication: Faster than Fast-LTS: Robust Regression and Outlier Detection with DC Programming

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270460)