Large-Scale Linear Programming

WA03.1 Improved Approaches to Stochastic Optimization via Integrative Population Analysis Fred Glover, John M. Mulvey --- Univ. of CO, Sch. of Bus., CP 419, Boulder, CO 80309-0419, (fred.glover@colorado.edu)

Integrative population analysis introduces an adaptive design for¨ mapping problems to groups. We focus on processes embodied in¨ parameter based definitions of regionality. Applied to a practical¨ financial application, we succeed in generating regions and¨ representations that give 90-95% improvement in computational¨ efficiency over traditional approaches.

WA03.2 Ordering Sparse Matrices for Interior Point Linear Programming Edward E. Rothberg, Bruce Hendrickson --- Silicon Graphics, Inc., 2011 N Shoreline Blvd., MS 580, Mountain View, CA 94049 , (rothberg@sgi.com)

We evaluate several fill-reducing heuristics for ordering the¨ normal equations arising in interior-point methods for LP. We find¨ that several classes of large-scale LP problems a new method, based¨ on the nested dissection heuristics, provides significantly better¨ orderings than the most commonly used method, minimum degree.

WA03.3 Using LOQO to Solve Large Scale Convex Programs Robert J. Vanderbei, Bing Yang --- Princeton Univ., Dept. of Civ. Eng. & OR, Princeton, NJ 08544 , (rvdb@princeton.edu)

We describe how to use LOQO to solve large scale convex programming¨ problems. The main challenge is to develop a connection between the¨ solver and a problem generator. We explain how we interfaced a¨ solver, LOQO, with AMPL as well as with the standard input format¨ developed by Conn, et al.