Hi, Please ignore my previous mail. I found the bug. It is with the Matlab interface, not with glpk itself.
Ali. On Mon, Nov 17, 2008 at 3:28 PM, Ali Ozer Ercan <[EMAIL PROTECTED]> wrote: > Hi, > > I am using GLPK with matlab (with the glpkmex extension) and I found a > bug. Please extract the attached tarball and run glpk_fail_example.m > in Matlab (you need glpkmex, the interface to Matlab). It is a simple > routing optimization. The first plot shows the possible links between > nodes and the numbers (in blue) on the links show the capacity in > Mbps. The flows are 18Mbps so the routes cannot use the links with > 12Mbps capacity. This translates to an upper bound constraint in GLPK. > I solve the problem with both GLPK and Matlab's 'bintprog'. The > bintprog gives the right solution (last 3 plots) while the solution > given by GLPK (plots 2, 3 and 4) do use the links with 12 Mbps > capacity, which means the UB constaint is violated. This is also > detected by the lines 42--45 of solve_opt.m. The warning is executed > if the ub constraint is violated, and the exit status of glpk is 5. > > I don't know if this is a problem of GLPK or glpkmex, but since > glpkmex is just a Matlab interface to GLPK and it is GLPK who is > performing the optimization, I thought this is probably GLPK's bug. > > Best, > > Ali Ercan. > > -- > +++++++++++++++++++++++++++++++++++ > Ali Ozer Ercan > Post Doctoral Researcher > Berkeley Wireless Research Center > 2108 Allston Way, Suite 200 > Berkeley, CA 94704-1302 USA > http://www.eecs.berkeley.edu/~aliercan > +++++++++++++++++++++++++++++++++++ > -- +++++++++++++++++++++++++++++++++++ Ali Ozer Ercan Post Doctoral Researcher Berkeley Wireless Research Center 2108 Allston Way, Suite 200 Berkeley, CA 94704-1302 USA http://www.eecs.berkeley.edu/~aliercan +++++++++++++++++++++++++++++++++++ _______________________________________________ Bug-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-glpk
