package ch.idsia.scenarios;
import ch.idsia.tools.EvaluationOptions;
import ch.idsia.tools.CmdLineOptions;
import ch.idsia.ai.Evolvable;
import ch.idsia.ai.ea.ES;
import ch.idsia.ai.tasks.MultiSeedProgressTask;
import ch.idsia.ai.agents.ai.LargeMLPAgent;
import ch.idsia.ai.agents.RegisterableAgent;
import ch.idsia.ai.agents.Agent;
import wox.serial.Easy;
/**
* Created by IntelliJ IDEA.
* User: julian
* Date: May 24, 2009
* Time: 1:18:44 AM
*/
public class EvolveMultiSeed {
final static int generations = 100;
final static int populationSize = 100;
public static void main(String[] args) {
EvaluationOptions options = new CmdLineOptions(new String[0]);
options.setMaxAttempts(1);
options.setPauseWorld(true);
Evolvable initial = new LargeMLPAgent();
if (args.length > 0) {
initial = (Evolvable) RegisterableAgent.load (args[0]);
//RegisterableAgent.registerAgent ((Agent) initial);
}
RegisterableAgent.registerAgent ((Agent) initial);
options.setMaxFPS(true);
options.setVisualization(false);
//Task task = new ProgressTask(options);
MultiSeedProgressTask task = new MultiSeedProgressTask(options);
task.setNumberOfSeeds(3);
task.setStartingSeed(0);
ES es = new ES (task, initial, populationSize);
System.out.println("Evolving " + initial + " with task " + task);
for (int gen = 0; gen < generations; gen++) {
//task.setStartingSeed((int)(Math.random () * Integer.MAX_VALUE));
es.nextGeneration();
double bestResult = es.getBestFitnesses()[0];
System.out.println("Generation " + gen + " best " + bestResult);
options.setVisualization(gen % 5 == 0 || bestResult > 4000);
options.setMaxFPS(true);
Agent a = (Agent) es.getBests()[0];
a.setName(((Agent)initial).getName() + gen);
RegisterableAgent.registerAgent(a);
double result = task.evaluate(a)[0];
options.setVisualization(false);
options.setMaxFPS(true);
Easy.save (es.getBests()[0], "evolved.xml");
if (result > 4000) {
break; //finished
}
}
}
}