genetix v1.3.1

Genetix

A simple library to test genetic algorithms

Installation

npm install -S genetix

Description

With Genetix you can create simple genetic algorithms. Its flexibility opens a lot of doors, and lets you customize every aspect of the algorithm.

Keep in mind that this is a library meant for testing and not a library meant for actual production/calculus code. It is SLOW and unoptimized.

Genetix uses 3 main classes Gene, Genome and Population.

tl;dr

To use Genetix you first initialize a Population object with a fitness function and generators. generators is a list of functions that return Genes.

Example:

let Genetics = require('genetix');

// Create generator function
let generator = function(current_genes){
	let int = Math.floor(Math.random() * 30) + 1;
	return new Genetics.Gene(int);
}

// All generators are the same
let generators = new Array(4).fill(generator);

// Create population with generators and fitness function
let pop = new Genetics.Population(generators, function(genes){
	// retrieve gene values
	let vals = genes.map(e => e.value());

	// return inverse of 30 - function (so that fitness is an increasing curve)
	let inverse = 30 - (vals[0] + 2*vals[1] + 3*vals[2] + 4*vals[3]);
	return 1/inverse;
});

// Evolve until a satisfying genome is found (30 is the target, 1000 max generation)
let last_gen = pop.auto(function(genes){
	let vals = genes.map(e => e.value());
	return (vals[0] + 2*vals[1] + 3*vals[2] + 4*vals[3]);
}, 30, 1000);

fitness takes as argument the genes of every Genome and expects a number as output.

Performance

You should read this at the end, but I know you won't read the whole doc, so I'm putting it here.

You should test many different mutation_rates (between 0 and 1 in probability) to find optimal execution time.

As an example, when running examples/mutation_rate.js, I get 1min50 sec on average and no exact match for mutation_rate = 0.05 and 400ms for mutation_rate=0.2 with an exact match (running 2.3GHz i7);

Try running the example on your computer.

Concepts

Population

A population is like a generation (actually, a generation in this case is composed of a population). It contains individuals, with a maximum number of them, capable of reproduction.

These indiivuals reproduce to combine their genes and go towards a given goal, measured by the fitness function.

The population class is defined as

Constructor

new Population(generators[], fitness(), [crossover()], [params], [population])

Methods

.fit()

Executes fitting of all Genomes in population

.breed(was_fit)

Breeds and returns new pre-filled Population. was_fit is a boolean that specifies if the Population was already fit, or needs fitting. Default is false;

.auto(pheno(), target, [max_generations], [gen_callback)

Automatically breeds populations until one reaches the given target.

Genome

A Genome is practially the same thing as an Individual. It is composed of a list of Genes.

Constructor

new Genome(genes=[], fitness, crossover, mutation)

Methods

.fit()

Fits the Genome using fitness function given in constructor or Population.

.crossover(genome2, generators)

Crosses over curernt Genome with given Genome using given crossover in constructor or Population. Using mutation_rate randomly decides to add a random mutation to genes using given generators.

.pheno(fct)

Renders phenotype using given Pheno function.

Gene

A gene is the basic component of a Genome.

Constructor

new Gene(value, type=null)

Methods

.value()

Returns value for gene. If function, executes and returns result.