RANDMOD

Linear programming model randomizer

RANDMOD is designed to enable random variations in a resident LP, whose matrix file was previously generated, such as by MODLER or GAMS, or was read a library, such as NETLIB). The purpose in doing so is to test either the performance of an algorithm or a response to the model's sensitivity to variations. The following figure represents a job flow for statistical analysis, putting the function of RANDMOD into perspective. (The "base matrix" is from the modeling language or the library.)

     .==============.
     |  Data files  |    .============.
     |--------------|    | LP library |
     | Model library|    |(eg, NETLIB)|
     `--------------'    `------------'
            |                  |
            |         .==================.
    .--- MODLER ------| Base matrix file |
    |                 `------------------'
    |                          |
    |       .-------------------------------------.
    |       |           .=============.           |
    |    RANDMOD -------| Matrix file |------ Optimizer ---.
    |       |           `-------------'           |        |
    |       |                  |         .===============. |
    |       |               ANALYZE -----| Solution file | |
    |       |                  |         `---------------' |
    |       |            .============.                    |
    |       `------------| Statistics |--------------------U
    `--------------------| Collection |
                         `------------'
                               |
             Statistical Analysis/Reporting Package

The experiment can repeat entrances into RANDMOD with the base matrix (or the last matrix) to generate new random variations. Each matrix file is an instance, or random realization, of the base matrix. The optimizer solves each one for some specified number of cases, or with a dynamic stopping rule, such as use of sequential analysis.

Need more info?

Visit the RANDMOD page (maintained by Harvey J. Greenberg) or contact

Prof. Harvey J. Greenberg
Math Dept - Campus Box 170
University of Colorado at Denver
PO Box 173364
Denver, CO 80217-3364
phone: 303-556-8464 fax: 303-556-8550
hgreenberg@carbon.cudenver.edu

Reference:

A Primer for RANDMOD: A System for Randomizing Modifications to an Instance of a Linear Program, 1995. Technical Report, Mathematics Department, University of Colorado at Denver, Denver, CO.

RANDMOD: A System for Modifications to an Instance of a Linear Program, ORSA Journal on Computing 3:2 (1991) 173-175.

User's Guide

[ Optimization Software Guide | OTC Home Page | NEOS Server | NEOS Guide ]