/// <summary>
/// Invoked when this page is about to be displayed in a Frame.
/// </summary>
/// <param name="e">Event data that describes how this page was reached.
/// This parameter is typically used to configure the page.</param>
protected async override void OnNavigatedTo(NavigationEventArgs e)
{
_captureManager = await new MediaCaptureWrapperFactory().Create();
capturePreview.Source = _captureManager.MediaCapture;
// _captureManager.StartPreview();
var picturesLibrary = await StorageLibrary.GetLibraryAsync(KnownLibraryId.Pictures);
await Task.Run(() =>
{
_classifier = NeuralNetFactory.CreateNetWithPredefinedWeights(picturesLibrary.SaveFolder.Path + @"\weights.txt", 13, 6);
});
}
public void TestHogOnAllNewDataWithPerformanceAndConfidence(string dataPath)
{
var folds = 5;
for (int i = 0; i < folds; i++)
{
var allData = Reading.ReadAllFiles(dataPath, folds);
var testData = allData[i];
allData.RemoveAt(i);
var data = Reading.MergeReadings(allData);
Link.RenewalFactor = 0.000003; // maybe decrease
Link.Step = 0.05;
var sut = NeuralNetFactory.Create(2,
10,
data.InputValues[0].Count,
data.ExpectedResults[0].Count);
var iterations = sut.TrainAdaptiveWithConfidenceAndPerformance(data.InputValues, data.ExpectedResults, -0.00005, 150000, testData, i, dataPath);
}
}
public static void Test(string[] args)
{
// Read training set
string trainFile = args[0];
DataFrame df = DataFrame.Read(trainFile, sep: '\t');
foreach (string s in df.Headers())
{
Console.WriteLine(s);
}
double[] x = df["x"].ToDouble();
double[] y = df["y"].ToDouble();
// Prepare training set
int nTrain = x.Length;
double[][] xTrain = new double[nTrain][];
double[][] yTrain = new double[nTrain][];
for (int i = 0; i < nTrain; i++)
{
xTrain[i] = new double[] { x[i] };
yTrain[i] = new double[] { y[i] };
}
// Set up model
string activation = "Relu";
int nHiddenLayers = 0;
int nNeuronsPerLayer = 2;
int inputSize = xTrain[0].Length;
int outputSize = yTrain[0].Length;
double alpha = 0.5;
double eta = 0.15;
NeuralNet model = NeuralNetFactory.Get(activation, nHiddenLayers, nNeuronsPerLayer,
inputSize, outputSize, alpha, eta);
model.Fit(xTrain, yTrain);
// Predict
string predFile = args[1];
df = DataFrame.Read(predFile, sep: '\t');
foreach (string s in df.Headers())
{
Console.WriteLine(s);
}
//double[] xPred_ = df["x"].ToDouble();
//double[][] xPred = new double[xPred_.Length][];
//for (int i = 0; i < xPred_.Length; i++)
// xPred[i] = new double[1] { xPred_[i] };
double[][] xPred = df.SetAsDouble(new string[] { "x" });
double[] yPred = model.Predict(xPred);
// Output result
string resFile = args[2];
object[][] resData = yPred.Select(z => new object[] { z }).ToArray();
CsvManager.Write(resData, resFile, sep: '\t');
//model.BackPropagation();
string ss = "";
ss += "";
//string type = "Standard";
//Scaler scaler = Scaler.Create(type);
//scaler.Fit();
}
