[Clp] CLP versus GLPK versus CPLEX

[John Forrest]

Victor,

As a small part of a dissertation, there are clues to how Clp works (at
least for the dual algorithm) in -

http://ubdok.uni-paderborn.de/servlets/DerivateServlet/Derivate-5257/disserta.pdf

You may find it helpful

John Forrest
On Mon, 2011-08-01 at 14:08 -0400, Victor Miller wrote:
> I have a number of fairly large LP problems for which I just want to
> find any feasible solution.  I've given these problems to GLPK, CLP
> and CPLEX "out of the box" (i.e. not specifying any parameters, just
> asking for the default behavior).  Although all of them do find the
> feasible vectors that I desired, only CLP produces "nice" vectors.
> For instance one of the problems has 32768 variables divided into two
> sets of 16384.  CLP produces a vector for which only 3 out of the
> 16384 variables in the first set are nonzero, and all the nonzero
> values for variables in the second set are 1.0 and all of the nonzero
> indices occur in consecutive chunks of 256.  My question is what is
> the default solution strategy for CLP (and if anyone knows for GLPK
> and CPLEX).  Clearly, by serendipity, CLP is finding some nice
> structure in my problem, so I'd like to uncover it.
> 
> Victor Miller
> 
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp