public double AskNeuralNetwork(double[] input)
{
Network.Network network = new Network.Network(_inputData[0].Length, _numberOfHiddenLayers, _networkTopology);
network.InitializeNetwork();
network.SetWeights(_archive[0]);
return network.CalculateOutputForOneExample(input);
}
private void ComputeFitness()
{
//for (int i = 0; i < _2archiveSize; i++)
//{
// Network.Network network = new Network.Network(_inputData[0].Length, _numberOfHiddenLayers, _networkTopology);
// network.InitializeNetwork();
// network.SetWeights(_archive[i]);
// _fitnessVector[i] = calculateMAEForAllExamples(network);
//}
Parallel.For(0, _2archiveSize, i =>
{
Network.Network network = new Network.Network(_inputData[0].Length, _numberOfHiddenLayers, _networkTopology);
network.InitializeNetwork();
network.SetWeights(_archive[i]);
_fitnessVector[i] = CalculateMaeForAllExamples(network);
});
}
private void SetTotalNumberOfWeights()
{
Network.Network network = new Network.Network(_inputData[0].Length, _numberOfHiddenLayers, _networkTopology);
_totalNumberOfWeights = network.TotalNumberOfWeights;
}
double CalculateMaeForAllExamples(Network.Network network)
{
//double error = _inputData.Select((t, i) => Math.Abs(_desiredOutput[i] - network.CalculateOutputForOneExample(t))).Sum();
double error = 0.0;
for (int i = 0; i < _desiredOutput.Length; i++)
{
double networkOutput = network.CalculateOutputForOneExample(_inputData[i]);
double absError = Math.Abs(_desiredOutput[i] - networkOutput);
if ((_desiredOutput[i] > 0 && networkOutput > 0) || (_desiredOutput[i] < 0 && networkOutput < 0))
{
error += absError;
}
else
{
error += 10*absError;
}
}
return error / _inputData.Length;
}