Global Maximization of Functions with Lots of Local Maxima Using Simulated Annealing or Other Snazzy Methods
- To: MathGroup at yoda.physics.unc.edu
- Subject: Global Maximization of Functions with Lots of Local Maxima Using Simulated Annealing or Other Snazzy Methods
- From: "J. Seth" <SCHandler at uh.edu>
- Date: Sun, 26 Dec 1993 13:04:15 -0600 (CST)
I have a function of several variables that I am attempting to maximize. Because the function often has several local maxima I have to "cheat" and use inside information about the function. I also find that the FindMinimum built-in procedure in Mathematica is rather sluggish. Here are some ideas that I have considered. 1. Use something called simulated annealing that "Numerical Recipes in C: The Art of Scientific Computing" suggests is good at finding the global maximum of functions with lots of local maxima. Does anyone have any experience with simulated annealing in Mathematica? Better yet, does anyone know of a simulated annealing package. 2. Figure out some way to speed up FindMinimum by sacrificing accuracy or precision. I would consider using MathLink and C, but Mathematica for Windows ver 2.2 does not yet support it. Any help would be deeply appeciated. Professor Seth J. Chandler, University of Houston