kono
7db5a2af3b
- only setting the seed of RNG once at the beginning and not reseeding it afterwards. Deep copying the initial AntWorld to use as blueprint for resetting the world instead of reseeding and creating pesudo random again. Reseeding the RNG has influence action selecting to always choose the same trajectory. - instance of is used to determine if policy has epsilon or not and the view will adopt to this, only showing epsilon slider if policy has epsilon
42 lines
1.2 KiB
Java
42 lines
1.2 KiB
Java
package core.policy;
|
|
|
|
import core.RNG;
|
|
import lombok.Getter;
|
|
import lombok.Setter;
|
|
|
|
import java.util.Map;
|
|
|
|
/**
|
|
* To prevent the agent from getting stuck only using the "best" action
|
|
* according to the current learning history, this policy
|
|
* will take random action with the probability of epsilon.
|
|
* (random action space includes the best action as well)
|
|
*
|
|
* @param <A> Discrete Action Enum
|
|
*/
|
|
public class EpsilonGreedyPolicy<A extends Enum> implements EpsilonPolicy<A>{
|
|
@Setter
|
|
@Getter
|
|
private float epsilon;
|
|
private RandomPolicy<A> randomPolicy;
|
|
private GreedyPolicy<A> greedyPolicy;
|
|
|
|
public EpsilonGreedyPolicy(float epsilon){
|
|
this.epsilon = epsilon;
|
|
randomPolicy = new RandomPolicy<>();
|
|
greedyPolicy = new GreedyPolicy<>();
|
|
}
|
|
|
|
@Override
|
|
public A chooseAction(Map<A, Double> actionValues) {
|
|
System.out.println("current epsilon " + epsilon);
|
|
if(RNG.getRandom().nextFloat() < epsilon){
|
|
// Take random action
|
|
return randomPolicy.chooseAction(actionValues);
|
|
}else{
|
|
// Take the action with the highest value
|
|
return greedyPolicy.chooseAction(actionValues);
|
|
}
|
|
}
|
|
}
|