public void StartGridSearch(List <List <Face> > faces)
{
var stopwatch = new Stopwatch();
stopwatch.Start();
var date = DateTime.Now.ToString("yyyy-dd-M-HH-mm-ss");
var networkLearningInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Learning, 12);
var networkLearningOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Learning, 15);
var networkValidationInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Validation, 12);
var networkValidationOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Validation, 15);
var networkTestingInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Testing, 12);
var networkTestingOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Testing, 15);
var learningSet = NetworkHelper.NormaliseDataSet(networkLearningInput, networkLearningOutput);
var validationSet = NetworkHelper.NormaliseDataSet(networkValidationInput, networkValidationOutput);
var testingSet = NetworkHelper.NormaliseDataSet(networkTestingInput, networkTestingOutput);
Parallel.ForEach(_momentum, (x) => MainCalculationLoop(_learningRate, x, _hiddenLayersCount, _bias, _neuronsCount, learningSet, validationSet, testingSet, faces, date, stopwatch.Elapsed));
stopwatch.Stop();
}
private void MainCalculationLoop(double[] _learningRate, double _momentum, int[] _hiddenLayersCount, bool[] _bias, int[] _neuronsCount, INeuralDataSet learningSet,
INeuralDataSet validationSet, INeuralDataSet testingSet, List <List <Face> > faces, string date, TimeSpan timeFromStart)
{
foreach (var learningRate in _learningRate)
{
foreach (var neuronsCount in _neuronsCount)
{
foreach (var hiddenLayersCount in _hiddenLayersCount)
{
foreach (var bias in _bias)
{
var inputData = new InputClass(learningRate, _momentum, hiddenLayersCount, neuronsCount, bias, 15, new ActivationSigmoid(), 40000);
NetworkHelper.LearnNetwork(learningSet, testingSet, faces[0][0].features.Count, 15, inputData, validationSet);
Task.Factory.StartNew(() =>
XmlFileWriter.WriteDataToFile("GridSearch" + date + ".xml", inputData.LearningError, inputData.ValidationError, inputData.TestingError, inputData.ElapsedTime, inputData.IterationsCount,
learningRate, _momentum, hiddenLayersCount, neuronsCount, bias, timeFromStart)
);
}
}
}
}
}
public void PerformCalculation(InputClass inputData, List <List <Face> > faces, string calcStartDate, TimeSpan timeFromStart, bool[] activeFeatures = null)
{
Console.WriteLine("Szykuje dane zbioru uczacego");
var networkLearningInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Learning, 12, activeFeatures);
var networkLearningOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Learning, 15);
var networkValidationInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Validation, 12, activeFeatures);
var networkValidationOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Validation, 15);
var networkTestingInput = NetworkHelper.CreateNetworkInputDataSet(faces, 12, 5, DataSetType.Testing, 12, activeFeatures);
var networkTestingOutput = NetworkHelper.CreateNetworkOutputDataSet(faces, 12, 5, DataSetType.Testing, 15);
var learningSet = NetworkHelper.NormaliseDataSet(networkLearningInput, networkLearningOutput);
var validationSet = NetworkHelper.NormaliseDataSet(networkValidationInput, networkValidationOutput);
var testingSet = NetworkHelper.NormaliseDataSet(networkTestingInput, networkTestingOutput);
NetworkHelper.LearnNetwork(learningSet, testingSet, faces[0][0].features.Count, this.PeopleCount, inputData, validationSet);
Task.Factory.StartNew(() =>
XmlFileWriter.WriteDataToFile("Genetic" + calcStartDate + ".xml", inputData.LearningError, inputData.ValidationError, inputData.TestingError, inputData.ElapsedTime, inputData.IterationsCount,
inputData.LearningFactor, inputData.Momentum, inputData.HiddenLayers, inputData.HiddenNeurons, inputData.Bias, timeFromStart, activeFeatures)
);
}
