/** When the state has widely varying Q-values. */
public boolean giveAdvice(BasicRLPacMan student, MOVE _choice, MOVE _advice) {
double[] qvalues = student.getQValues();
double gap = Stats.max(qvalues) - Stats.min(qvalues);
// System.out.println(gap);
boolean uncertain = (gap < threshold);
if (uncertain) {
left--;
lastStudentActionCorrect = false;
return true;
} else lastStudentActionCorrect = true;
return false;
}
/** When the state has widely varying Q-values, and the student doesn't take the advice action. */
public boolean giveAdvice(BasicRLPacMan teacher, MOVE choice, MOVE advice, Game game) {
double[] qvalues = teacher.getQValues();
double gap = Stats.max(qvalues) - Stats.min(qvalues);
boolean important = (gap > threshold);
if (important) {
boolean mistake = (choice != advice);
if (mistake && step >= m) {
left--;
step = 0;
return true;
}
}
step++;
return false;
}