public void PredictRandomFromTrainingSet()
{
if (trained)
{
double[] testInput = dataSets [testValueIndex].Values;
double[] results = network.Compute(testInput);
double maxPrediction = -1;
int loc = -1;
for (int i = 0; i < results.Length; i++)
{
if (maxPrediction < results [i] && (i - (numOutputs / 2)) != 0)
{
maxPrediction = results [i];
loc = i;
}
}
Debug.Log((loc - (numOutputs / 2)));
}
}
// Update is called once per frame
void Update()
{
// Read input
var inputData = new double[_numInputParameters];
for (int i = 0; i < _numInputParameters; i++)
{
inputData[i] = inputs[i].get();
}
// Compute output
outputData = _network.Compute(inputData);
output.set(outputData[0], outputData[1]);
}
public double[] Predict(double[] values)
{
if (trained)
{
double[] results = network.Compute(values);
//Debug.Log (results [1]);
return(new double[] { (results [0] * 60 - 30) }); //, results [1] });
}
else
{
return new double[] { 0 }
}; //, 0 };
}
void TrainNetwork()
{
while (true)
{
if (network == null)
{
network = new NeuralNetwork.NetworkModels.Network(numInputs, hiddenLayer, numOutputs);
network.LearnRate = learningRate;
network.Momentum = momentum;
}
Debug.Log("Network is training, this can take a while...");
if (trainType == TrainingType.Epoch)
{
network.Train(dataSets, epochs);
}
else
{
network.Train(dataSets, minimumError);
}
Debug.Log("Training complete!");
trained = true;
break;
}
}
