[FlopCpp] Using FLOPC++ with GLPK MathProg examples

[Greg Geiselhart]

Hi,

I've been investigating the feasibility of converting GLPK MathProg problems
to FLOPC++. To understand how to convert a MathProg model to FLOPC++, I
started with a simple comparison of FLOPC++ examples/transport.cpp to GLPK
examples/transport.mod. In executing both examples, I seem to get
conflicting optimal solutions:

 

*       glpsol calculates an optimal solution of 156.375

*       FLOPC++ calculates an optimal solution of 156.15

 

In looking at transport.cpp, it occurred to me that the problem statement
may not exactly match the GLPK transport.mod (in transport.cpp,
transportation links do not seem to be defined for Seattle->Topeka and San
Diego->New York). To test weather this accounted for the solution
differences, I modified transport.cpp to add those links (the code is shown
below). When the modified transport problem is executed, the optimal
solution is calculated as 153.675. 

 

Can anyone offer an explanation for the difference in calculated optimal
solutions? It would seem a direct translation of the GLPK transport.mod
would require the additional links added in the code below. But this change
seems to increase the discrepancy with GLPK.

 

Thanks in advance for any insights

 

 

int main() {

    MP_model::getDefaultModel().setSolver(new OsiClpSolverInterface);

    enum  {seattle, sandiego, numS};

    enum  {newyork, chicago, topeka,numD};

 

    MP_set S(numS);          // Sources 

    MP_set D(numD);          // Destinations 

    MP_subset<2> Link(S,D);  // Transportation links (sparse subset of S*D)

 

    Link.insert(seattle,newyork);

    Link.insert(seattle,chicago);

    Link.insert(seattle,topeka);

    Link.insert(sandiego,newyork);

    Link.insert(sandiego,chicago);

    Link.insert(sandiego,topeka);

 

    MP_data SUPPLY(S);

    MP_data DEMAND(D);

 

    SUPPLY(seattle)=350;  SUPPLY(sandiego)=600;

    DEMAND(newyork)=325;  DEMAND(chicago)=300;  DEMAND(topeka)=275;

 

    MP_data COST(Link);

 

    COST(Link(seattle,newyork)) = 2.5;

    COST(Link(seattle,chicago)) = 1.7;

    COST(Link(seattle,topeka)) = 1.8;

    COST(Link(sandiego,newyork))= 2.5;

    COST(Link(sandiego,chicago))= 1.8;

    COST(Link(sandiego,topeka)) = 1.4;

 

    COST(Link) = 90 * COST(Link) / 1000.0;

 

    MP_variable x(Link);

    x.display("...");

 

    MP_constraint supply(S);

    MP_constraint demand(D);

 

    supply.display("...");

 

    supply(S) =  sum( Link(S,D), x(Link) ) <= SUPPLY(S);

    demand(D) =  sum( Link(S,D), x(Link) ) >= DEMAND(D);

 

    cout<<"Here"<<endl;

 

    minimize( sum(Link, COST(Link)*x(Link)) );

/*

    assert(MP_model::getDefaultModel()->getNumRows()==5);

    assert(MP_model::getDefaultModel()->getNumCols()==4);

    assert(MP_model::getDefaultModel()->getNumElements()==8);

    assert(MP_model::getDefaultModel()->getObjValue()>=156.14 &&
MP_model::getDefaultModel()->getObjValue()<=156.16);

*/

    x.display("Optimal solution:");

    supply.display("Supply dual solution");

    cout<<"Test transport passed."<<endl;

}

 

 

Greg Geiselhart

email: geiselha at optonline.net

gmail: geiselha at gmail.com

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/flopcpp/attachments/20070130/80900147/attachment.html