Monday, May 20
4:50-6:50 PM
Esquinalt
MS13
Linear Algebra for Interior Point Methods
This minisymposium will focus on computational linear algebra for interior point methods.
Interior point methods are now the preferred approach to large scale structured linear
programming problems. Almost all the computational work associated with these methods is
finding the Newton step, which requires solution of a system of linear equations with special
structure. These equations can be written in so called ``KKT'' form and also as weighted least
squares problems, and they can be highly ill conditioned. The speakers in this minisymposium
will discuss recent advances in the development of efficient and stable algorithms for
determining the Newton step.
Organizer: Stephen A. Vavasis
Cornell University
- Stable Solution of Weighted Least Squares for Near-Degenerate Linear
Pro-gramming Problems
- Patricia Hough, Cornell University and S. Vavaris, Organizer
- Preconditioners and the Iterative Solution of the Linearized KKT-Systems in
Linear Programming
- Florian Jarre, Universit„t Wurzburg, Germany; and Roland Freund,
Massachusetts
Insitute of Technology
- Solution of KKT Systems within OSL's Barrier Algorithm
- Michael Saunders, Stanford University; and J. Tomlin, IBM Almaden Research Center
- Finite Precision Effects in Interior-Point Methods
- Stephen Wright, Argonne National Laboratory
LMH, 3/15/96