Optimizers, part 1

Post date: Jun 19, 2009 12:42:41 PM

The question of which optimizer to use for this project has become very difficult. It seems that everyone who has an expertise using a particular algorithm thinks that algorithm is superior to all others, however when less experienced users attempt to apply them, the results are not as good.

Simulated Annealing

Pros: implemented in Scipy, global optimizer, easy to understand

Cons: can be difficult to find robust tuning parameters

Particle Swarm

Pros: advocated by BJ

Cons: not implemented in mature python library.