Examples of EGreedyPolicy

net.javlov.policy.EGreedyPolicy
Epsilon-greedy (e-greedy) policy. Chooses action with maximum Q-value with probability 1 - e, and a (uniform) random action with probability e. This means that the greedy action will be chosen with probability 1 - e + e/(nr of actions). @author Matthijs Snel

Examples of net.javlov.policy.EGreedyPolicy

  
  protected Agent makeAgent(List<? extends Option> optionPool) {
    qf = TabularQFunction.getInstance(optionPool.size());
    SarsaAgent a = new QLearningAgent(qf, 1, optionPool);
    a.setLearnRate(new DecayingLearningRate(1, optionPool.size(), 0.8));
    Policy pi = new EGreedyPolicy(qf, 0.1, optionPool);
    a.setPolicy(pi);
    a.setSMDPMode(false);
    return a;
  }

View Full Code Here

Examples of net.javlov.policy.EGreedyPolicy

  protected SarsaAgent makeAgent(List<? extends Option> optionPool) {
    qf = TabularQFunction.getInstance(optionPool.size());
    SarsaAgent a = new QLearningAgent(qf, 1, optionPool);
    //a.setLearnRate(new DecayingLearningRate(1, optionPool.size(), 0.8));
    a.setLearnRate( new FixedLearningRate(0.2) );
    Policy pi = new EGreedyPolicy(qf, 0.05, optionPool);
    a.setPolicy(pi);
    a.setSMDPMode(true);
    a.setLearnStateValueFunction(false);
    a.setInterruptOptions(false);
    return a;

View Full Code Here

TOP

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.