[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-glpk] Problem Diagnostics

From: Andrew Makhorin
Subject: Re: [Help-glpk] Problem Diagnostics
Date: Mon, 21 Jan 2013 19:36:01 +0400


> I've modified the test generator so it no longer has dependencies on
> local libraries.  I've also expanded the precision used in the
> MathProg data file to reduce truncation related errors. It's attached
> along w/ a Makefile and a new driver script.
> The problem manifests itself if part of the first row is populated by
> zeros.  The README and other files in the tarball give the details.
> If I'm mistaken in my expectation of what the simplex solver should
> produce I'd greatly appreciate a pointer to an explanation of my
> error.  I can think of other ways to solve my problem.
> Any suggestions of further work I can do are most welcome.  Also
> please let me know if I've omitted something.  I've been working on
> this for several weeks and it's gotten to be a bit of a blur.  I've
> chased a lot of dead ends trying to understand why I was getting bad
> results on certain problem instances.

I generated .dat files with your tst1 script and then run bug.mod using
some of data files--glpsol successfully solved all the instances.

Sorry, may be I missed something, but what is the issue?

> For example, is there currently a way to obtain the determinant of the
> non-basic matrix in glpsol?  I found lots of instances of "determine"
> with "egrep determin", but none of "determinant".  

Matrix determinants are generally not used in practical computations. To
estimate the error in solution of a linear system and, in particular, to
decide if a given matrix is close to a singular one with the working
precision you may use its condition number. AFAIK, GSL has a condition
estimator for dense matrices (glpk doesn't have).

> If there is not, would an optional dependency on the Gnu Scientific
> Library be acceptable?  That has code to compute the determinant from
> an LU decomposition that I might be able to adapt to work in GLPK to
> monitor the evolution of the solution.

Could you please explain what exactly do you need to do?

Andrew Makhorin

reply via email to

[Prev in Thread] Current Thread [Next in Thread]