JSGA

JavaScript Genetic Algorithm

JSGA is an ES6 genetic algorithm library.

Installation

• Bower: bower install jsga-feklee

• Yarn: yarn add jsga-feklee

Or you can download jsga.js from the GitHub repo.

Usage

JSGA operates on chromosomes --- lists of numbers. It generates successive generations of chromosomes by mixing the chromosomes of randomly selected parents, and by randomly mutating chromosomes. It evaluates each individual by using a user-provided fitness function. Only the fittest individuals from each generation are carried over to the next.

Start by creating a JSGA object:

let algorithm = JSGA({
    length: 10,
    radix: 10,
    fitness: function(individual) {
        return individual.reduce((p, c) => p + c); // the fitness is the sum of all the values
    },
    size: 100,
    children: 4,
    mutationRate: 0.05,
    crossovers: 1
});

Notes on the parameters:

• length is the length of the chromosome.
• radix is the 'base' used to encode the data. Each number in the chromosome will have a range from 0 to radix.
• fitness is the fitness function used to evaluate individuals. It is passed a chromosome (array of numbers) and expects a number to be returned. JSGA will try to maximize the fitness function.
• size is the number of individuals in the population. This must be an even number.
• seed is used to fill the first individuals of the initial population. The remaining individuals are random.
• children is the number of offsping each pair of parents will produce. The default is 4.
• mutationRate is the rate of random mutations, in mutations/number. That is, for the default mutation rate of 0.05, around one in every twenty numbers in each chromosome will be randomly mutated.
• crossovers is the number of times to randomly cross over the parents chromosomes. This is the primary source of variation. The default is 1.

To run the algorithm, simply call the run method with the number of generations to run the algorithm. Pass -1 to make the algorithm run forever. run is a generator which yields an object containing information about the current generation.

for (let generation of algorithm.run(50)) {
    console.log(`Generation ${generation.generation}`);
    console.log(`Array of individuals: ${generation.population}`);
    console.log(`Best individual: ${generation.best.params}`);
    console.log(`Best individual's fitness: ${generation.best.fitness}`);
}