binet-ga.js

const fibonacci = lucas(0, 1);
const dataset = Array(79)
	.fill()
	.map(_ => fibonacci.next().value);
let generation = 0;
let population = Array(10000)
	.fill()
	.map(_ => Array(4).fill().map(_ => Math.random() * 10 - 5))
	.map(evalFitness);

console.log(generation, population[0], makeSequence(population[0]));

// Generation iterator
(function iterate() {
	population = population
		.sort((a, b) => a.fitness - b.fitness)
		.slice(0, 1000)
		.flatMap(chromosome => Array(10).fill(chromosome))
		.map(chromosome => chromosome.map(gene => {
			if (Math.random() < 0.0001) return Math.random() * 10 - 5;
			if (Math.random() < 0.001) return -gene;
			return gene + mutation(chromosome.fitness);
		}))
		.map(evalFitness);
	++generation;
	console.log(generation, population[0], makeSequence(population[0]));
	if(population[0].fitness > 100) setTimeout(iterate, 0);
})();

// Lucas numbers generator
function* lucas (a, b) {
	while (true) {
		yield a;
		[ a, b ] = [ b, a + b ];
	}
}

// Evaluate the fitness of a given chromosome
function evalFitness (chromosome) {
	chromosome.fitness = dataset.reduce((deviation, actual, n) => {
		const [ c1, b1, c2, b2 ] = chromosome;
		const computed = c1 * b1 ** n + c2 * b2 ** n;
		return deviation + Math.abs(actual - computed);
	}, 0);
	return chromosome;
}

// Generate number sequence from a chromosome
function makeSequence ([ c1, b1, c2, b2 ]) {
	return Array(20)
		.fill()
		.map((_, n) => Math.round( c1 * b1 ** n + c2 * b2 ** n ));
}

// Generate mutation value
function mutation (fitness) {
	return (Math.random() - 0.5) * fitness / 1e16;
}