void FixedUpdate()
{
int bestAction = expert.EvaluateAction(rMemory, threshold, samples);
int lastAction = learner.CallAction(bestAction);
expert.AddToMemory(CLAB);
float predictionError = expert.PredictionError(memory, CLAB, lastAction);
//Debug.Log("prediction Error" + predictionError);
kga.AddToErrorMemory(predictionError);
float meanError = kga.MeanError(eMemory, samples);
//Debug.Log("mean Error" + meanError);
float metaError = kga.MetaM(eMemory, samples, offset);
//Debug.Log("meta Error" + meanError);
float reward = kga.Reward(meanError, metaError);
Debug.Log("reward" + reward * 10);
expert.AddToRewardMemory(CLAB, reward * 10, lastAction);
}
void FixedUpdate()
{
int bestAction = expert.EvaluateAction(rMemory, threshold, samples, availableAcions);
sensorReading = ReadSensors(sensors);
predictionModel.SensoryActionDataAtT.Add(double.Parse(sensorReading.ToString() + bestAction.ToString()));
double prediction = expert.MakePredition(double.Parse(sensorReading.ToString() + bestAction.ToString()));
availableAcions[bestAction].DoAction(CLAB);
//now at t+1
sensorReading = ReadSensors(sensors);
predictionModel.SensoryDataAtTPlus1.Add(sensorReading);
int lastAction = bestAction;
float predictionError = expert.PredictionError(prediction, sensorReading);
//Debug.Log("Prediction Error:" + predictionError);
//Debug.Log("prediction Error" + predictionError);
kga.AddToErrorMemory(predictionError);
float meanError = kga.MeanError(eMemory, samples);
//Debug.Log("mean Error" + meanError);
float metaError = kga.MetaM(eMemory, samples, MetaMemOffset);
//Debug.Log("meta Error" + meanError);
float reward = kga.Reward(meanError, metaError);
Debug.Log("Reward: " + reward * 10 + " To Action: " + lastAction);
expert.AddToRewardMemory(reward * 10, lastAction);
CullErrorMemory(eMemory, MaxMemorySamples);
CullRewardMemory(rMemory, MaxMemorySamples);
CullPredictionMemory(predictionModel.SensoryDataAtTPlus1, MaxMemorySamples);
CullPredictionMemory(predictionModel.SensoryActionDataAtT, MaxMemorySamples);
UpdateThreshold();
// Debug.Log("New Threshold: " + threshold);
}
