dinosaurs eat everybody

for all your Dave Schwantes related needs

Darwinning, My Ruby Gem for Genetic Algorithms

Well, this writeup is a bit late. I wanted to write about Darwinning right after I launched it but here we are.

So, what is Darwinning?

Darwinning is a Ruby Gem built to help people write genetic algorithms. I got into genetic algorithms while taking a machine learning course during my masters and I've had a casual interest in them ever since. I even wrote a comic about them! I have also been enjoying Ruby lately, so I wanted to learn how to build a Gem. This project was started to satisfy both of those interests.

I also played with the Ruby unit testing framework, RSpec, for the first time with this project. The test coverage for Darwinning isn't great, but at least it's a start.

So, how do I use Darwinning?

Darwinning is totally available via RubyGems so you can install it the same way you would install almost any Gem:

gem install darwinning

Once it's installed you can make use of the Population, Organism, and Gene classes to build your own genetic algorithm based programs.

Here's an dumb example of how you might use Darwinning to solve a pointless problem:

Let's say for some reason you need a set of 3 number that add up to 15. This is a strange problem to have, but let's solve it anyway.

class Triple < Darwinning::Organism

  @name = "Triple"
  @genes = [
      Darwinning::Gene.new("first digit", (0..9)),
      Darwinning::Gene.new("second digit", (0..9)),
      Darwinning::Gene.new("third digit", (0..9))
    ]

  def fitness
    # Try to get the sum of the 3 digits to add up to 15
    (genotypes.inject{ |sum, x| sum + x } - 15).abs
  end
end 

p = Darwinning::Population.new(Triple, 10, 0, 0.1, 100)
p.evolve!

p.best_member.nice_print # prints the member representing the solution

This code declares an organism class that inherits from Darwinning's Organism parent class to represent solutions. Then we create a population of these solution organisms and evolve the population until a solution meets the fitness threshold or the generation limit is met.

So, what else can I do with it?

If any of this sounds remotely interesting to you, please check out the project on Github. I'm always happy to take good pull requests if other people want to contribute. I haven't been very active lately because I've been spending a lot of time on this website and my wedding website, but I do keep an eye on the project.