Annals of Operations Research 47(1993)443--482 443
A simplified global convergence proof of the
affine scaling algorithm*
R.D.C. Monteiro
Department of Systems and Industrial Engineering, University of Arizona,
Tucson, AZ 85721, USA
T. Tsuchiya
The Institute of Statistical Mathematics, 4-6-7 Minami-Azabu, Minato-Ku,
Tokyo 106, Japan
Y. Wang
Department of Systems and Industrial Engineering, University of Arizona,
Tucson, AZ 85721, USA
This paper presents a simplified and self-contained global convergence proof for
the affine scaling algorithm applied to degenerate linear programming problems.
Convergence of the sequence of dual estimates to the center of the optimal diial face
is also proven. In addition, we give a sharp rate of convergence result for the sequence
of objective function values. All these results are proved with respect to the long step
version of the affine scaling algorithm in which we move a fraction A, where
A G {0,2/3], of the step to the boundary of the feasible region.
1. Introduction
The affine scaling algorithm, introduced by Dikin [6] in 1967, is one of the
simplest and most efficient interior point algorithms for solving linear program-
ming (LP) problems. Because of the theoretical and practical importance of the
affine scaling algorithm, there are a number of papers which study its global and
local convergence [3, 6-8, 11, 16, 19-24] and the behavior of its associated con-
tinuous trajectories [2, 4, 16, 25]. As in the simplex algorithm, the analysis of the
affine scaling algorithm for (primal) degenerate LP problems is much harder than
for (primal) nondegenerate LP problems.
*This research was supported by the National Science Foundation (NSF) under Grant No. DDM-
9109404 and the Overseas Research Scholars of the Ministry of Education, Science and Culture of
Japan.
© J-C. Baltzer AG, Science Publishers