genome.js

genome.js is a Javascript to help build insane genetics algorithms in a few minutes.

Concept

General terms

• Population: a subset of the possible solutions to the problem (ie. subset of chromosomes)

• Chromosome: a specific solution to the problem

• Gene: a value defining a chromosome

Specific terms

• Blueprint: a schema defining the structure of every gene (number and possible values) in a chromosome.

Installation (via NPM)

npm install --save genome.js

Documentation

Population

|Methods|Return type|Description| |--|--|--| | constructor(size: number, blueprint: Blueprint) |Population|Create a population with size chromosomes using the blueprint| |setFitnessCalculation(fitnessCalculation: any)|null|Set the fitness calculation function. It should return a number value corresponding to the fitness of a chromosome.| | setStopAt(fitness: number) |null|Stop the process once a chromosome reaches AT LEAST fitness value on its fitness.| | setMutationRate(mutationRate: number) |null|Set the mutation rate value. It should be between 0 (no mutation at all) and 1 (every chromosome will be mutated)| | setCutOff(cutOff: number) |null|Set the cut off value. It should be between 0 (no chromosome will be removed) and 1 (every chromosome will be removed)| | run(rounds: number = 1) |null|Run the process rounds times.| | getGenerationNumber() |number|Return the current round number.| | getBestChromosome() |Chromosome|Return the best chromosome.|

Chromosome

|Methods|Return type|Description| |--|--|--| | getGenes() |Gene[]|Return the genes of the chromosome.| | getFitness() |Gene[]|Return the fitness of the chromosome.|

Gene

|Methods|Return type|Description| |--|--|--| | get() |number|Return the allele (value) of the gene.|

Blueprint

|Methods|Return type|Description| |--|--|--| | constructor() |Blueprint|Create a new Blueprint.| | add(factor: number, times: number = 1) |null|Define a property into the blueprint. The factor is used when you get back the allele (value) of a gene (ex: a gene created with add(26) will return a number between 0 and 25). You can add times a property by setting the times parameter.|

GenoveEvent

|Methods|Return type|Description| |--|--|--| | static on(eventType: GenomeEventType, callback: any) |null|STATIC Run the callback function when the event eventType is trigger.|

Events

|Name|Description| |--|--| |GENOME_EVENT_POPULATION_CREATED|Trigger when all chromosomes are initialized| |GENOME_EVENT_GENERATION_BEGIN|Trigger when a new generation is processed| |GENOME_EVENT_GENERATION_END|Trigger when a generation is done processing| |GENOME_EVENT_GENERATION_FINISH|Trigger when the all processing is done (rounds limit or fitness limit)|

Example

/*
* This example is based on the "infinite monkey theorem" (https://en.wikipedia.org/wiki/Infinite_monkey_theorem)
*
* The algorithm tries to reproduce a specific text input, here "helloworldhowareyoutoday" in a minimum rounds.
*/

// Importing all the dependencies
import { Population, Blueprint, Gene, Chromosome, GenomeEvent, GenomeEventType } from 'genome.js';

// Defining the string to reproduce
const answer = 'helloworldhowareyoutoday';

// We create a blueprint to represent the data structure of a chromosome
const blueprint = new Blueprint();
// Our chromosomes will have 'answer.length' genes between 0 and 26 (not included), so that each gene can represent one letter of the alphabet
blueprint.add(26, answer.length);

// We generate a population of 500 chromosomes using our blueprint
const population = new Population(500, blueprint);

// Just some basic configurations
population.setMutationRate(0.01);
population.setCutOff(0.5);
population.setStopAt(100); // We stop the processing when a chromosome reach AT LEAST 100 on his fitness

// We define now the function that calculate the fitness of every chromosome on each generation
// Be sure to never return 0 (cause a bug, WIP)
population.setFitnessCalculation((genes: Gene[]) => {
    let sum = 1; // Avoid to have 0 on fitness

    for (let i = 0; i < genes.length; i += 1) {
        const charCode = answer.charCodeAt(i) - 97;
        const geneCharCode = Math.floor(genes[i].get());
        // If the gene value is corresponding with the answer letter at the same location, then increment 'sum'
        if (charCode === geneCharCode) {
        sum += 1;
    }
}

// Basically a percent of correct genes' values
return (sum / (genes.length + 1)) * 100;
});

// We wait for a generation to end, and we display the best chromosome fitness into the console
GenomeEvent.on(GenomeEventType.GENOME_EVENT_GENERATION_END, (chromosomes: Chromosome[]) => {
    const bestChromosome = chromosomes[0];
    console.log(`Generation ${population.getGenerationNumber()}: ${bestChromosome.getFitness()}`);
});

// Once the process in finished (when a chromosome reach the fitness limit or the process has reach the round limit), we display the string contained in its genes
GenomeEvent.on(GenomeEventType.GENOME_EVENT_GENERATION_FINISH, (chromosomes: Chromosome[]) => {
    let finalString = '';
    const bestChromosome = chromosomes[0];
    bestChromosome.getGenes().map((gene: Gene) => {
        finalString += String.fromCharCode(gene.get() + 97);
    });
    console.log(`Result (fitness: ${bestChromosome.getFitness()}): ${finalString}`);
});

// We process the algorithm throught 500 rounds (more options comming soon)
population.run(500);