Quantitative genetic algorithm

note that this is research, there are no guarantees

This is based on work by Jakub Otwinowski and Colin H. LaMont.

install

using Pkg
Pkg.add("https://github.com/jotwin/qga")

How to use

using QGA
minimize(f, mu, sd, s)

• f: An objective function which takes a vector as input
• mu: mean of initial distribution (vector)
• sd: standard deviation of initial distribution (vector)
• s: target entropy

Keyword arguments

• Fmintarget: stop when target value of objective reached (default -Inf)
• sdmin: stop when target value of standard deviation of fitness reached (default 0)
• MaxEvals: stop after (number) iterations (default 10000*dim)
• trace: show info every (number) iterations (default Inf)
• n: number of unique variants (default exp2(s+1))

Output is a dictionary with :halt describing reason for stopping. :DupFit means duplicate fitnesses were found and is generally a failure to converge.

example

dim = 5
fd = 2 .^(1:dim)
fellipse(x) = sum(fd.*x.^2)
minimize(fellipse, zeros(dim), ones(dim), 5.0, sdmin = 0.0, n = 512,
            FminTarget = 1e-8, trace = 100*dim)