Title of presentation:
A Stable Iterative Method for Linear Programming
at seminar for Computational Mathematics at the University of Waterloo
authors:
Maria Gonzalez-Lima \thanks{Research supported by Universidad Sim\'on
Bol\'{\i}var and Conicit (project G97000592), Venezuela. E-mail
{mgl@cesma.usb.ve} }
\and
Hua Wei\thanks{ Research supported by The Natural Sciences and
Engineering Research Council of Canada and Bell Canada. E-mail
h3wei@math.uwaterloo.ca}
\and
\href{http://orion.math.uwaterloo.ca/~hwolkowi/}{Henry Wolkowicz}
\thanks{Research supported by The Natural Sciences and Engineering
Research Council of Canada. E-mail {hwolkowicz@uwaterloo.ca}
abstract
We present a new primal-dual interior/exterior-point method
for linear programming. We use a simple preprocessing step to
eliminate both the primal and dual feasibility equations. We then
apply an iterative method, within an inexact Newton framework,
directly on the linearized equations. We present numerical examples
where roundoff error causes problems for the Normal Equation approach.
The numerical tests show that our method takes direct advantage of
sparsity and stability of the data.