static void CNN(List <DenseVector[]> Data)//畳み込みニューラルネットワーク
{
//MNISTの読み込み
MNistImageLoader imgLoader = MNistImageLoader.Load("train-images.idx3-ubyte");
MNistLabelLoader lLoader = MNistLabelLoader.Load("train-labels.idx1-ubyte");
//////////データ用意
Random rnd = new Random();
List <DenseVector[]> xds = new List <DenseVector[]>();
for (int i = 0; i < 10000000; i++)
{
xds.Add(new DenseVector[2] {
new DenseVector(1), new DenseVector(1)
});
double a = rnd.NextDouble();
xds[i][0][0] = a; xds[i][1][0] = 5 * a * a;
}
List <DenseVector[]> Tests = new List <DenseVector[]>();
for (int i = 0; i < 100; i++)
{
Tests.Add(new DenseVector[2] {
new DenseVector(1), new DenseVector(1)
});
double a = rnd.NextDouble();
Tests[i][0][0] = a; Tests[i][1][0] = a * a;
}
//////////データ用意終
UseNN UseCNN = new UseNN();
UseCNN.Add(new int[] { imgLoader.numberOfRows *imgLoader.numberOfColumns, 100, 100, 100, 100, 100, 100, 20, 20, 9 }
, new int[] { 0, 1, 2, 1, 2, 1, 2, 0, 0, 0 });
}
static void Main(string[] args)
{
//MNISTの読み込み
MNistImageLoader imgLoader = MNistImageLoader.Load("train-images.idx3-ubyte");
MNistLabelLoader lLoader = MNistLabelLoader.Load("train-labels.idx1-ubyte");
Console.WriteLine(imgLoader.numberOfImages);
Console.WriteLine(lLoader.numberOfItems);
int n = 3375;
byte[] image = imgLoader.bitmapList[n];
byte label = lLoader.labelList[n];
for (int i = 0; i < imgLoader.numberOfRows; i++)
{
for (int j = 0; j < imgLoader.numberOfColumns; j++)
{
if (image[i * imgLoader.numberOfRows + j] >= 255)
{
Console.Write("#");
}
else
{
Console.Write(" ");
}
}
Console.WriteLine();
}
Console.WriteLine(label);
Console.ReadLine();
}