ssusnic / machine-learning-flappy-bird Goto Github PK

    var pop = [];
    for (var i = 0; i < GA.Population.length; i++) {//  GA = new GeneticAlgorithm(max_units, winners_count)
        pop.push(GA.Population[i].toJSON());
    }
    fs.writeFileSync('./population.json', JSON.stringify(pop));
    console.log("saved");

if (fs.existsSync('./population.json')){
    var d = fs.readFileSync('./population.json', 'utf8');
    var pop = JSON.parse(d);
    for (var i = 0; i < pop.length; i++) {
        var unit = synaptic.Network.fromJSON(pop[i]);
        unit.index = i;
        unit.fitness = 0;
        unit.score = 0;
        unit.isWinner = false;
        GA.Population[i] = unit;
        console.log('unit has been read');
    }
}
Score(GA)

crossOver : function(parentA, parentB) {
		var cutPoint = this.random(0, parentA.neurons.length-1);
		for (var i = cutPoint; i < parentA.neurons.length; i++){
			//what is the bias part. what does it mean
			var biasFromParentA = parentA.neurons[i]['bias'];
			parentA.neurons[i]['bias'] = parentB.neurons[i]['bias'];
			parentB.neurons[i]['bias'] = biasFromParentA;
		}

		return this.random(0, 1) == 1 ? parentA : parentB;
	},

mutation : function (offspring){
		for (var i = 0; i < offspring.neurons.length; i++){
			offspring.neurons[i]['bias'] = this.mutate(offspring.neurons[i]['bias']);
		}
		
		//is connections a keyword?
		for (var i = 0; i < offspring.connections.length; i++){
			offspring.connections[i]['weight'] = this.mutate(offspring.connections[i]['weight']);
		}
		
		return offspring;
	},

activateBrain: function(bird, target){
			var horizontalDistance = this.normalize(target.x,700) *this.SCALE_FACTOR;
			var verticalDistance = this.normalize(bird.y - target.y,800) * this.SCALE_FACTOR;	
			var inputs = [horizontalDistance,verticalDistance];
			//not sure what activate is
			var outputs = this.Population[bird.index].activate(inputs);
			if(outputs[0] > 0.5) bird.flap();
	},