The algorithm used by MIPS is fully documented in Appendix A in the User's Manual, especially in section A.4.
Many things contribute to the speed of NLP solvers and there can be significant variation depending on the size and other characteristics of the problem being solved. So it's hard to say why MIPS might be the fastest in your tests. I often find PDIPM, KNITRO and sometimes Ipopt to be faster on large problems, and MINOPF is very hard to beat on small problems. Regarding case3375wp, I'm not sure what it is, but there seems to be something very numerically challenging about that case. MIPS and PDIPM fail, but KNITRO is able to solve it with no problems. -- Ray Zimmerman Senior Research Associate 419A Warren Hall, Cornell University, Ithaca, NY 14853 phone: (607) 255-9645 On Apr 10, 2013, at 9:49 AM, Peter Raush <[email protected]> wrote: > Dear Prof. Zimerman, > > I was trying to compare the convergence time of different solver in Matpower. > It turns out that MIPS solver is the fastest one (especially in comparing to > Matlab fmincon) when using Matpower test cases. Also, I found out that MIPS > solver doesn't converge for the largest test case 3375wp. > > Could you please let me know why MIPS solver converges faster and is there a > place where I can find more information about it? > > Thank you very much for your support, > Peter
