In this page it is provided the software that has been used to solve the problem of finding the Maximum Expected Utility (MEU) in a Limited Memory Influence Diagram (Limid) by Variable Elimination with pareto pruning. The method is described in this arxiv document (in this arxiv version the algorithm is called by its old name, lve, but they are essentially the same).
Academic non-profit use is permitted as long as the proper citation (shown below) is included in the work, but modifications, redistributions, for-profit usage, and/or non-academic usage are not allowed (please write to me if you need a for-profit and/or non-academic use). The program is provided without warranty of any kind, either expressed or implied. The complete risk of its use is with you.
If you make use of it in a publication, I kindly ask you to cite the following paper:
@inproceedings{MauaNIPS2011,
author={Denis D. Mau\'a and Cassio P. de Campos},
title={Solving Decision Problems with Limited Information},
booktitle={Neural Information Processing Systems Conference (NIPS)},
pages={to appear},
year={2011},
}
Current, the software is available here as a linux binary file named cif2mip.exe and some MATLAB scripts that work as a wrapper for the solver. The Limid has to be specified using the BNT package. Within MATLAB, one calls the function
[meu,time,maxdimen,maxparetosize] = mpu(bnet,tl,verbose,exepath,fname)The parameters are explained inside the mpu.m file. In short, bnet is a Limid (as defined using the Limid type within BNT package), tl is the number of seconds of a time limit, verbose is zero or one (self-explained), exepath gives the full path of the cif2mip.exe file, and fname is the name of the temporary file used by the procedure. All arguments are optional except for bnet. The output contains meu (expected utility), time (spent to run), maxdimen (maximum number of dimensions of any potential during computations), maxparetosize (maximum number of elements in any set propagated during computations).