void TrainWorks()
{
using (var ip = new ImageProvider())
{
var imageProvider = ip.ImageStream();
List <(Vector <double>, int)> tmp = imageProvider
.Select(i => new { pixels = i.SampledPixels, i.Label })
.Select(i => (Vector <double> .Build.DenseOfEnumerable(i.pixels), i.Label)).ToList();
var X = tmp.Select(i => i.Item1).ToList();
var Y = tmp.Select(i => i.Item2).ToList();
imageProvider.Select(i => i.Label).ToList();
Random r = new Random(1);
var net = new MatrixNetwork(784);
net.AddLayer(200, r);
net.AddLayer(10, r);
net.Train(X, Y);
int performance = 0;
int total = 0;
foreach (var el in X.Zip(Y, (input, expected) => new { input, expected }))
{
total++;
var result = net.Predict(el.input, out List <double> output);
if (el.expected == result)
{
performance++;
}
}
Assert.That(performance, Is.InRange(6000, total));
}
}
public void InitialFileClamped0to256()
{
using (ImageProvider i = new ImageProvider())
{
foreach (var el in i.ImageStream())
{
foreach (var pixel in el.Pixels())
{
Assert.That(pixel, Is.InRange(0, 255));
}
}
}
}
public void ValidateDatabaseInput()
{
using (ImageProvider i = new ImageProvider())
{
var stream = i.ImageStream().GetEnumerator();
stream.MoveNext();
Assert.That(stream.Current.Label, Is.EqualTo(5));
stream.MoveNext();
Assert.That(stream.Current.Label, Is.EqualTo(0));
stream.MoveNext();
Assert.That(stream.Current.Label, Is.EqualTo(4));
stream.MoveNext();
Assert.That(stream.Current.Label, Is.EqualTo(1));
stream.MoveNext();
Assert.That(stream.Current.Label, Is.EqualTo(9));
}
}
