evo-js v0.2.5

evoJS

evoJS is a genetic and evolutionary algorithm tool, making it easy to incorporate evolutionary algorithms in your JavaScript.

Installation

To install run npm install evo-js.

To run on a browser, include evo.min.js in your scripts.

Basic Usage

The tool is used through a population object. Here is a simple example

    // Instantiate a population object
    var population = evo.population();

    var evaluateGenes = function (genes) {
      /* Use genes to return a score */
    }

    // Define a method to evaluate genes and return the fitness
    // 'genes' is an array of floats
    population.on('run', function(genes){
        return evaluateGenes(genes); // Supply this method to evaluate the genes
    });

    // Run with a stopping criteria
    population.run({generations: 10});

    // Get the best genes
    var result = population.bestGenes();

Pool Configuration

As an argument to evo.population a configuration object can be passed to override the defaults. Here are the defaults.

    var config = {
        genes: 200, // Number of genes for each member
        size: 100, //  Number of members in the gene population
        cross_rate: 0.05, // The frequency of gene "twists" in two parents genes
        mutate_rate: 0.05, // The frequency of mutations in a parent gene
        mutate_amount: 1.0, // The amount a mutated gene can deviate

        // Each generation is made from a mix of different breeding strategies
        // Ratios defines the ratio of each in the next generation
        ratios: {
            top:    0.25,    // Survivors from last generation
            mutate: 0.25,    // Created by randomly altering genes
            cross:  0.25,    // Created from crossing parents
            random: 0.10,    // Random survivors from last generation
            average:0.05,    // Created by averaging parents together
            fresh:  0.10     // new, totally random gene sets
        }
    });

    var population = evo.population(config);

Stopping conditions

As an argument to population.run a configuration object can be passed to defined the stopping conditions

    var config = {
        iterations: 1000, // How many times run will be called
        generations: 100, // How many generations will be run
        score: 10.0,      // Minimum score to be reached (higher score is better)
        // A while function can be supplied
        // when false is returned sim stops
        while: function(){
            return true;
        }

        // If true, a method will be attempted to stop when the score
        // stops changing across generations
        auto_run: false
    });

    population.run(config);

'Manual' simulations

While population provides the run method to run simulations, the gene population can be accessed manually for custom simulations.

    population = evo.population()

    while(population.generation < 100) {
        // Grab next genes
        var genes = population.nextGenes();

        //Evaluate genes
        var score = evaluateGenes(genes);

        //Report genes back to population
        population.report(genes, score);
    }

When the gene population is empty, a new generation will be created.

Member construction

You may provide a member constructor to automate gene evaluation, and use a more object oriented approach

    population = evo.population()

    // Supply a 'constructor' to use the genes
    population.on('member', function(genes){
        var member = {
            height: genes[0],
            width: genes[1]
        };
        return member;
    });

    // Run will now give your constructed member instead of a gene list
    population.on('run', function(member){
        return member.height + member.width;
    });