Nonlinear optimization problems often have many local solutions. Nonlinear solvers like CONOPT, IPOPT and KNITRO will normally return the first local solution that they find, even though it might be far away from the global optimum. The multistart algorithm in AIMMS can be used to find better solutions.
In this webinar, we discuss the multistart algorithm and demonstrate how it can be used. The multistart algorithm has been parallelized in AIMMS 4.9 to utilize multiple CPU cores. Among other things, we demonstrate that using parallel multistart can drastically reduce computation time.