Regularized Nonsmooth Newton Algorithms for Best Approximation
Abstract
We consider the problem of finding the best approximation point from a polyhedral set, and its applications, in particular to solving large-scale linear programs. The classical projection problem has many various and many applications. We study a regularized nonsmooth Newton type solution method where the Jacobian is singular; and we compare the computational performance to that of the classical projection method of Halperin-Lions-Wittmann-Bauschke (HLWB). We observe empirically that the regularized nonsmooth method significantly outperforms the HLWB method. However, the HLWB has a convergence guarantee while the nonsmooth method is not monotonic and does not guarantee convergence due in part to singularity of the generalized Jacobian. Our application to solving large-scale linear programs uses a parametrized projection problem. This leads to a \emph{stepping stone external path following} algorithm. Other applications are finding triangles from branch and bound methods, and generalized constrained linear least squares. We include scaling methods that improve the efficiency and robustness.
Cite
@article{arxiv.2212.13182,
title = {Regularized Nonsmooth Newton Algorithms for Best Approximation},
author = {Yair Censor and Walaa M. Moursi and Tyler Weames and Henry Wolkowicz},
journal= {arXiv preprint arXiv:2212.13182},
year = {2023}
}
Comments
38 pages, 7 tables, 8 figures