/// <summary>
/// Implement the Q Learning update rule using experience replay
/// </summary>
private void UpdateKnowledge()
{
var mini_batch = GenerateMiniBatch();
var trainingData = new List <Tuple <State, double[]> >();
var actions = Enum.GetValues(typeof(ActionType)).Cast <ActionType>();
foreach (var experience in mini_batch)
{
var targetValues = new Dictionary <ActionType, double>();
var estimatedValues = Q[experience.From];
foreach (var action in actions)
{
targetValues[action] = QUpdate(experience, action, estimatedValues[action]);
}
trainingData.Add(new Tuple <State, double[]>(experience.From, targetValues.Values.ToArray()));
}
Q.Train(trainingData);
OnTrainingComplete?.Invoke(this, new OnTrainingCompleteArgs());
}
public async Task __DoneTraining__()
{
OnTrainingComplete?.Invoke();
}
public void __DoneTraining__()
{
OnTrainingComplete?.Invoke();
}
