[Coin-lpsolver] CLP and QUADOBJ
Matthew Guthaus
mguthaus at eecs.umich.edu
Sun Aug 21 19:56:10 EDT 2005
Hi all,
I've been successfully using CLP to solve QP problems. However, I had
a QP problem that seemed to be taking a while so I decided to
writeMPS and run it in CPLEX for comparison. CPLEX complained when
the MPS file specifies the full Q matrix with:
CPLEX Error 5011: Duplicate entry in 'objective' for pair 'C0000012'
and 'C0000014'.
It is assuming that the matrix is symmetric and I should only specify
one entry (C12,C14) or (C14,C12). My question is, is this part of the
MPS specification or is this an assumption made in CPLEX? Is CLP
making the wrong assumption? Is it because CLP handles arbitrary Q
matrices and CPLEX can only handle semi-definite?
The other issue (which is actually a CPLEX problem, but maybe someone
will know!) is that CPLEX then claims my matrix is not semi-definite
when it surely is...
Q = (x2-x3)^2 + (x2-x4)^2 + (x2-x5)^2 + (x3-x4)^2 + (x3-x5)^2 + (x4-
x5)^2
Q = [ 0 0 0 0 0 0; 0 0 0 0 0 0 ; 0 0 3 -1 -1 -1 ; 0 0 -1 3 -1 -1 ; 0
0 -1 -1 3 -1 ; 0 0 -1 -1 -1 3 ]
octave:8> eig(Q)'
ans =
0.00000 0.00000 0.00000 4.00000 4.00000 4.00000
Thanks,
Matt
More information about the Clp
mailing list