Genetic algoritm

This is genetic algoritm implementation written in Elixir. It uses Flow library to achieve high concurrency in time-consuming tasks such as calculating fitness or crossovering chromosomes.

Rulette select is used to select chromosomes for crossover, it’s based on selecting chromosomes with chance coresponding to their fitness.

Examples of usage can be viewed here
One test is finding specific gene pattern, second searches for RNP expression of given number (e.g. 2 11 * for 22)

Installation

mix do deps.get, compile

Building docs

mix docs

Built docs can be accessed in doc/ folder or viewed online here.

Running tests and code style check

mix test --cover && mix credo --strict

Usage

iex -S mix

This command load Elixir’s REPL with project modules loaded. Example:

# Define fitness function:
fitness_f = fn genes ->
  desired = [%Gene{v: 1}, %Gene{v: 2}, %Gene{v: 3}, %Gene{v: 4}, %Gene{v: 5}]
    count = Enum.count(genes)

    fitting =
      desired
      |> Enum.zip(genes)
      |> Enum.reduce(0, fn {x, y}, acc -> if(x == y, do: 1, else: 0) + acc end)

    fitting / count
end

# Set algoritm options:
settings = [
  gene_values: [{Enum.to_list(1..5), 1.0}],
  length: 5,
  min_fitness: 1.0
]

# And call [`GeneticAlgorithm.find_solution/2`](GeneticAlgorithm.html#find_solution/2):
GeneticAlgorithm.find_solution(fitness_f, settings)

Result:

%{
  generations: 2,
  most_fitting: [
    %Chromosome{
      fitness: 1.0,
      genes: [
        %Gene{v: 1},
        %Gene{v: 2},
        %Gene{v: 3},
        %Gene{v: 4},
        %Gene{v: 5}
      ]
    }, 
    %Chromosome{
      fitness: 1.0,
      genes: [
        %Gene{v: 1}, 
        %Gene{v: 2},
        %Gene{v: 3},
        %Gene{v: 4},
        %Gene{v: 5}
      ]
    },
    %Chromosome{
      fitness: 0.8,
      genes: [
        %Gene{v: 1},
        %Gene{v: 1},
        %Gene{v: 3},
        %Gene{v: 4},
        %Gene{v: 5}
      ]
    }
  ]
}