Skip to content
Sleepingwell edited this page Dec 7, 2011 · 2 revisions

This code provides a couple of non-linear optimisation functions. Specifically, a Nelder-Mead based simulated annealing function copied straight from Numerical Recipes, and a Charged Particle search, based on:

https://svn-d1.mpi-inf.mpg.de/AG1/MultiCoreLab/papers/KavehTalatahari10%20-%20Charged%20System%20Search.pdf

A copy of that paper can be found in the root directory of this project (... I hope this is not breaking any copyright).

I got a little excited at the time I was doing this and set the code up such that the template ParticleSearcher should serve as a reasonable base class for a whole range of Swarm Intelligence algorithms (see http://en.wikipedia.org/wiki/Swarm_intelligence). As many of them seem to have a very similar structure, with the main differences being the distance metrics and attraction functions. I think further factoring of ParticleSearcher is possible to allow for specification of just these aspects of the algorithm via policies.

Clone this wiki locally