private Data ReadMnistData(string pathToLabels, string pathToImages)
{
var trainingImageLabels = FileReaderMNIST.LoadImagesAndLables(pathToLabels, pathToImages).ToList();
var trainingSampleSize = trainingImageLabels.Count;
var firstTestCase = trainingImageLabels.First().AsDouble();
var lineNb = firstTestCase.GetLength(0);
var colNb = firstTestCase.GetLength(1);
var trainingInputs = new double[lineNb * colNb, trainingSampleSize];
var trainingOutputs = new double[1, trainingSampleSize];
for (int tstNb = 0; tstNb < trainingSampleSize; tstNb++)
{
var currentSample = trainingImageLabels[tstNb];
for (int i = 0; i < lineNb; i++)
{
for (int j = 0; j < colNb; j++)
{
trainingInputs[i * lineNb + j, tstNb] = currentSample.Image[i, j];
}
}
trainingOutputs[0, tstNb] = trainingImageLabels[tstNb].Label;
}
var trainingData = new Data(trainingInputs, trainingOutputs);
return(trainingData);
}
public MnistDataSource(string imagesGzPath, string labelsGzPath)
{
var mnistData = FileReaderMNIST.LoadImagesAndLables(labelsGzPath, imagesGzPath);
_dataPoints = mnistData.Select(testCase =>
new DataPoint {
Value = ConvertImage(testCase.Image), Label = ConvertLabel(testCase.Label)
}).ToList();
}
private void Form1_Load(object sender, EventArgs e)
{
var data = FileReaderMNIST.LoadImagesAndLables(
"train-labels-idx1-ubyte.gz",
"train-images-idx3-ubyte.gz");
traindata = data;
var data2 = FileReaderMNIST.LoadImagesAndLables(
"t10k-labels-idx1-ubyte.gz",
"t10k-images-idx3-ubyte.gz");
testdata = data2;
}
static void Main(string[] args)
{
var data = FileReaderMNIST.LoadImagesAndLables(
"./data/train-labels-idx1-ubyte.gz",
"./data/train-images-idx3-ubyte.gz");
var training = data.Take(50000);
var test = FileReaderMNIST.LoadImagesAndLables(
"./data/t10k-labels-idx1-ubyte.gz",
"./data/t10k-images-idx3-ubyte.gz");
// var network = new Network(new[] { 784, 30, 10 });
// network.GradientDescent(training, 30, 10, 3.0, test);
Console.WriteLine("Creating a " + 784 + "-" + 30 + "-" + 10 + " neural network");
var network = new NeuralNetwork(new[] { 784, 100, 10 });
var trainingData = ConvertData(training);
var testData = ConvertData(test);
network.GradientDescent(trainingData, 30, 10, 3.0, testData);
}
// http://yann.lecun.com/exdb/mnist/
public static IEnumerable <(double[] image, int label)> GetMNISTHandwrittenNumbers(string labelFileName, string imageFileName)
{
var dataDirectory = $"{TrainingDataDir}/{MNISTHandwrittenNumbersDir}";
EnsureMNISTHandwrittenNumbersDataExists(dataDirectory);
var trainingDataSet = FileReaderMNIST.LoadImagesAndLables($"{dataDirectory}/{labelFileName}", $"{dataDirectory}/{imageFileName}");
foreach (var trainingData in trainingDataSet)
{
var trainingDataAsDoubleArray = new double[784];
var trainingDataAsDouble = trainingData.AsDouble();
for (var i = 0; i < 28; i++)
{
for (var j = 0; j < 28; j++)
{
trainingDataAsDoubleArray[j + (28 * i)] = trainingDataAsDouble[i, j];
}
}
yield return(trainingDataAsDoubleArray, trainingData.Label);
}
}
public MNISTmanager(string LPath, string IPath)
{
ClusterImageIndex = 0;
ClusterLabelIndex = 0;
data = Shuffle(FileReaderMNIST.LoadImagesAndLables(LPath, IPath).ToList());
}
public IEnumerator <TestCase> LoadFiles(string labelsZipFile, string imagesZipFile)
{
var imagesAndLabels = FileReaderMNIST.LoadImagesAndLables(labelsZipFile, imagesZipFile);
return(imagesAndLabels.GetEnumerator());
}
