Notes

Introduction

The SUS (http://en.wikipedia.org/wiki/Stochastic_universal_sampling) selection method is slow as treacle still, but I think I've got a decent implementation of the same idea in USER0 which isn't much slower than completely random selection.

Still needs a couple of updates, such as parameterising over the randomising function, and I think it could do with some refactoring. I'd like to use a design pattern to improve the population constructor for a start, as it's a bit unwieldy with so many parameters. Don't you have a book that talks about some design pattern that uses a class for initialiser so you can do things like:

population(popInit().popsize(100).mrate(0.2));

The only other major update I can think of is to add the alternative selection and mutation methods, as there's still a couple missing. It should compile fine under windows using MinGW, as there's no platform specific stuff, I think (not sure where timing functions fall on that score).

See the other pages on chromosome, population and sus_select for technical notes