CRPC-TR97777-S						Revised August 1998

Title: Numerical Comparisons of Path-Following Strategies for a Basic 
Interior-Point Method for Nonlinear Programming

Authors: M. Argaez, R.A. Tapia and L. Velazquez

Submitted September 1998; Available as Rice CAAM TR97-21

Abstract:
An important research activity in interior--point methodology for general
nonlinear programming is to determine effective path--following strategies
and its implementations using merit function technology. The objective of
this work is to present numerical comparisons of several globalization
strategies for the local interior--point Newton's method described by
El-Bakry, Tapia, Tsuchiya, and Zhang. We conduct a numerical
experimentation between three notions of proximity to the central path or
a relaxation of it, with three merit functions to obtain an optimal
solution. We present which strategy works best, as well as some
interesting comments about the rest of the strategies.

Keywords: interior-point method, path-following strategy, merit function.

Abbreviated Title: Numerical Comparisons.

------------------------------------------------------------------------------
M. Argaez
mar@caam.rice.edu
Departamento de Matematicas
Universidad del Valle

R.A. Tapia		L. Velazquez
rat@caam.rice.edu	leti@caam.rice.edu
Department of Computational and Applied Mathematics
Rice University