Return to Colloquia & Seminar listing
Multistage Stochastic Linear Programming on a Computational Grid
OptimizationSpeaker: | Prof. Jeff Linderoth, Lehigh University |
Location: | 2112 MSB |
Start time: | Fri, Apr 14 2006, 12:10PM |
We will describe an implementation of a nested-decomposition-based, multi-stage stochastic linear programming solver running on a computing platform known as a computational grid. The aim of the work is to create a "Monster Solver" capable of solving instances orders of magnitude larger than currently possible. We begin by reviewing the nested-decomposition algorithm, and then introduce the concept of a Computational Grid. Algorithmic mechanisms for dealing with the shortcomings of a Computational Grid are discussed. Computational results revealing the power of the approach and the platform will be given for an instance of a telecommunication capacity planning problem. This is joint work with Jierui Shen.