Logo image
Approximate structured optimization by cyclic block-coordinate descent
Technical documentation   Open access

Approximate structured optimization by cyclic block-coordinate descent

Jorge Villavicencio and Michael D. Grigoriadis
Rutgers University
1995
DOI:
https://doi.org/10.7282/t3-9etd-x391

Abstract

A uniform randomized exponential-potential block-coordinate descent method for the approximate solution of block-angular convex resource-sharing programs was analyzed in [5] and for the linear case in [14]. The former method is rendered deterministic by replacing its random block selection by arbitrary sweeps of its block coordinates, akin to classical implementations of Gauss-Seidel relaxation and coordinate descent in unconstrained optimization, recently used in concurrent network flows [15]. The general block-angular model consists of K disjoint convex compact sets ("blocks") and M nonnegative convex block-separable inequalities (coupling constraints"). It is shown that for linear coupling constraints and for a given but arbitrary relative accuracy " 2 (0; 1], the proposed derandomized algorithm runs in O(K lnM("2 + ln minfK; Mg) coordination steps or block optimizations, which is lower than all other existing bounds. It is also shown that this bound on coordination steps also applies to a reformulation of the above general nonlinear problem.
pdf
255191.86 kBDownloadView
Technical Documentation Open Access
url
Report an accessibility issueView
Please complete a content remediation request to report an accessibility issue with a library electronic resource, website, or service.

Metrics

178 File downloads
71 Record Views

Details

Logo image