Fast Simulated Annealing Optimiser
Uncategorized April 15th. 2008, 11:17pmhttp://asa-caltech.sourceforge.net/
I highly recommend this tool for your Optimisation purposes - very efficient code and a good example for learning how to implement Simulated Annealing Algorithm in C. What what inspires me is the approach the coder of the tool (Lester Ingber) has taken to implement the algorithm - well organised structure highlighting the prime benefits of C over Matlab.
April 16th, 2008 at 12:25 am
Ingber has a lot of nice code and done a lot of research on SA and ASA, indeed it’s a nice website. Also the current implementation of SA on the GA and Direct Search toolbox of Matlab is a good place to investigate.
April 16th, 2008 at 8:52 am
Yes matlab has the SA aswell but its terribly slow for high dimension problems. I suppose one can convert Ingber’s code into mex so that it can be run in matlab - atleast to overcome the speed issue.
April 16th, 2008 at 10:09 am
I dont think compiling Ingber’s code into a mex file would do much on performance. As i’ve seen recently, the performance diference bewteen mex-files and m-files that was common on versions older than R2006a, isn’t there anymore, they made a lot of performance improvements.
But i agree, i tryied Matlab SA once and it was kinda slow (most of the time i use GA or PSO), it all depends on the problem you are dealing with and the settings you use.
You can find PSO optimizers here:
http://www.adaptivebox.net/research/bookmark/psocodes_link.html
(i’ve used the one near the end of the list named “PSOt”, quite good).
April 16th, 2008 at 11:18 am
I meant using Ingber’s code in matlab rather than using matlab’s SA algorithm. I found the SA to be faster than GA and was primarily why I gave up on GA. Its problem dependent as you point out.