학술논문

A faster method for computing Karmarkar's projections for large number of constraints.
Document Type
Journal
Author
Spedicato, E. (I-BERG) AMS Author Profile; Nicolai, S. (I-BERG) AMS Author Profile; Yang, Z. (PRC-DUT-AM) AMS Author Profile
Source
Optimization. A Journal of Mathematical Programming and Operations Research (Optimization) (19970101), 40, no.~4, 343-350. ISSN: 0233-1934 (print).eISSN: 1029-4945.
Subject
65 Numerical analysis -- 65F Numerical linear algebra
  65F30 Other matrix algorithms

90 Operations research, mathematical programming -- 90C Mathematical programming
  90C08 Special problems of linear programming
Language
English
Abstract
Summary: ``In this paper we consider solving the linear system $AD^2_iA^\ssf Ty=AD^2_ic$, where $A$ is an $m\times n$ full row rank matrix, $m\leq n$, $y\in \bold R^m$, $c\in \bold R^n$, $D_i$ is diagonal with positive elements. Such a system characterizes the iteration of interior point methods and several problems in linear regression. We show that an equivalent system can be considered that has $n-m$ equations and variables, resulting in a significant reduction of the solution cost for large values of $m$.''