// Evolutionary Algorithm, simplified, for node, using `ChromosomeFloat` data structure
'use strict';
// * @license GPL v3
// * @package nodeo
// * @author V. M. Rivas <vrivas@ujaen.es>
// * From classic.js authored J. J. Merelo <jjmerelo@gmail.com>
// To avoid uncomprehensible radix complaint at charAt
/*jshint -W065 */
/*jshint smarttabs:true */
var utils = require('./Utils'),
CF = require('./chromosome-float'),
ops = require('./ops');
// Nodeo creates an evolutionary algorithms.
// `options` include
// * `population_size`
// * `fitness_func` that is the fitness used to evaluate all
// individuals. Can be either a function or a fitness object with `apply`
// * `tournament_size` using tournament selection. This is the number
// of individuals in each one
// * `pool_size` reproductive pool size
// * `chromosome_size` used to generate random chromosomes.
// * `minvalue` minimum value for floats
// * `maxvalue` minimum value for floats
// This function generates randomly a population which is used later
// on to start the evolutionary algorithms.
module.exports = function( options ) {
/*jshint validthis: true */
for ( var i in options ) {
this[i] = options[i];
}
Iif ( ! this.population_size ) {
return new Error ("0 population size");
}
Iif ( ! this.fitness_func ) {
return new Error ("No fitness func");
} else Eif ( typeof( this.fitness_func) === 'function' ) {
this.fitness_obj = new Fitness( options.fitness_func );
} else {
this.fitness_obj = this.fitness_func;
}
Eif ( !this.tournament_size ) {
this.tournament_size = 2;
}
Eif ( !this.pool_size ) {
this.pool_size = this.population_size - 2;
}
Iif ( !this.chromosome_size || isNaN(this.chromosome_size) ) {
throw "Chromosome size error";
}
Eif ( !this.minvalue || !this.maxvalue ) {
this.minvalue=0.0;
this.maxvalue=1.0;
}
this.population = [];
// console.log( this.fitness_obj.apply );
do {
var vector=[];
for( var i=0; i<this.chromosome_size; ++i ) {
vector[i]=Math.random()*(this.maxvalue-this.minvalue)+this.minvalue;
}
var chromosome = new CF.Chromosome( vector,
this.fitness_obj.apply( vector ) );
this.population.push( chromosome );
} while( this.population.length < this.population_size );
// Methods
this.tournament_selection = tournament_selection;
this.evaluation = evaluation;
this.generation= generation;
this.rank=rank;
};
// create fitness function object if it does not exist
function Fitness ( f ) {
/*jshint validthis: true */
this.apply = f;
}
// Selects a new population of size pool_size via comparing tournament_size chromosomes and taking the best
function tournament_selection( tournament_size, pool_size ) {
/*jshint validthis: true */
var pool = [];
Iif ( tournament_size <= 1 ) {
return new Error ("Tournament size too small");
}
do {
// var joust = [];
var best = this.population[ Math.floor(Math.random()*this.population.length) ] ;
for ( var i = 1; i < tournament_size; i ++) {
var another= this.population[ Math.floor(Math.random()*this.population.length) ];
if ( another.fitness > best.fitness) {
best = another;
}
}
pool.push( best );
} while (pool.length < pool_size );
return pool;
}
// Evaluates all the population not in cache
function evaluation( new_guys ) {
/*jshint validthis: true */
for (var i in new_guys) {
new_guys[i].fitness = this.fitness_obj.apply( new_guys[i].vector );
}
}
// sort population
function rank () {
/*jshint validthis: true */
var sorted_population = this.population.sort( function(a,b){ return b.fitness - a.fitness; } );
this.population = sorted_population;
}
// Single generation
function generation() {
/*jshint validthis: true */
var chosen = this.tournament_selection( this.tournament_size, this.pool_size);
this.rank(); // to get the best
var the_best = [this.population[0],this.population[1]];
var new_population = CF.reproduction( chosen);
this.evaluation(new_population);
this.population = the_best.concat( new_population );
this.rank(); // ranking twice????
}
|