mertsaner / EquilibriumScaling

C++ program that scales A, B, and C matrices of Linear, Integer and Mixed Integer Programs using the Equilibrium Scaling Method and returns Row and Column Scaling Factors 🎛

Equilibrium Scaling

Prerequisites

• To run the program firstly Gurobi C++ API should be ready. To install Gurobi Solver click here.
• The program requires the Eigen library to read large .lp files. Hence you need to include the Eigen library in the program. For Eigen library download instructions click here.

Usage

• To configure the program according to your needs, you need to edit the ScalingFactors.cpp file.
• In ScalingFactors.cpp edit lines 13-16 to feed the program with your LP/IP/MIP.
• Running the file will give you Row and Column Scaling Factors.
• Small_Investment_Model_presolved.lp, my.lp and trial.lp are three sample .lp files to test the program.

Notes

• Scaling.h is a header file that compares the result of the scaled program with the original program. To do that it uses a .xml file instead of .lp files like ScalingFactors.cpp.
• Optionally You can reach scaled A, B and C matrices, for that you need to uncomment certain parts of the code. While printing A, B, and C matrices, printing to the console is not recommended since it makes the program slow down. So it is better to write them to the file.
• You can reuse the scaling factors (R and C) or A, B and C depending on your own needs. For example, they can be fed into the program again and make the program solve the already scaled program. You can edit/update Scaling.h to compare the results of scaled and unscaled versions.