Dear Abouzar, Convergence failure for MIPS can be the result of an infeasible problem or it can simply be due to numerical issues, specifically near singular matrices in the computation of the Newton steps. The question of what characteristics of the problem result in such numerical issues is a much harder one and I'm afraid I don't have any good answers.
Unfortunately, because TSPOPF is based on the same algorithm, it usually has similar numerical difficulties when MIPS does. You might try one of the other solvers if they are available to you and the problem is not too large. The other option would be to give MIPS a different starting point. Unfortunately, this currently requires a modification to the code, since MIPS doesn't currently use the supplied starting point, but rather creates its own. I plan to make this optional in a future version. -- Ray Zimmerman Senior Research Associate B30 Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Oct 2, 2013, at 9:31 AM, Abouzar Estebsari <[email protected]> wrote: > > Dear subscribers > > I have a case to run OPF. PF can get converge but when I run OPF, due to > singular matrices, MIPS pops out some messages and the OPF cannot get > converge. > > I tried to install TSPOPF package but it had some errors and it could not be > run properly. I would like to have a general idea of what the most usual > reasons of MIPS failures are. > > Bests, > > Abouzar >
