cxjdavin / differentialevolution

Differential evolution implementation

About

An implementation of differential evolution based on Wikipedia (https://en.wikipedia.org/wiki/Differential_evolution).

Given a function f, the code will try to find the arguments which maximises the value of f.

There are 2 terminating conditions:

• When the maximum number of iterations has been reached
• When the best value of f (within a generation) improves less than a minimum threshold

Testing

The default test tries to maximise the function f(x,y) = -((x+5)^2 + 3 * \sqrt{|y+2| + 7}).

To optimise your own numerical function:

• Replace def f(args): ... with your own function
• Update argn
• [Optional] Modify tuning parameters such as N, DW, CR, max_iter and min_eps

Blog post

To learn more, read the blog post here: http://davinchoo.com/post/differentialevolution/