private void Form1_Load(object sender, EventArgs e)
{
Layer[] L1 = new Layer[2];
L1[0] = new Layer(784, Layer.ActivationFunction.ReLU);
L1[1] = new Layer(200, Layer.ActivationFunction.ReLU);
Layer[] L2 = new Layer[2];
L2[0] = new Layer(80, Layer.ActivationFunction.ReLU);
L2[1] = new Layer(10, Layer.ActivationFunction.Softmax);
LConnection[] C1 = new LConnection[1];
C1[0] = new LConnection(L1[0], L1[1]);
LConnection[] C2 = new LConnection[1];
C2[0] = new LConnection(L2[0], L2[1]);
NeuralNetwork[] NN = new NeuralNetwork[2];
NN[0] = new NeuralNetwork(L1, C1, 0.01);
NN[1] = new NeuralNetwork(L2, C2, 0.01);
NNConnection[] NNC = new NNConnection[1];
NNC[0] = new NNConnection(NN[0], NN[1]);
M = new Model(NN, NNC, Model.CostFunctions.CrossEntropy, 0.01);
D = new Dataset(784, 10);
D.LoadDataset("mnist_test.csv");
//string w = null;
// for (int i = 0; i < C[0].WeightBackMap.Length; i++)
// {
// w += C[0].WeightBackMap[i] + ",";
// }
//for (int i = 0; i < C[1].WeightBackMap.Length; i++)
//{
// w += (C[1].WeightBackMap[i]+25) + ",";
//}
//w += "\r\n";
//w += "\r\n";
//for (int i = 0; i < NN._WeightsBackMap.Length; i++)
//{
// w+=NN._WeightsBackMap[NN._WeightsBackMap.Length-i-1] +",";
//}
//MessageBox.Show(w);
Layer[] LTEST = new Layer[4];
LTEST[0] = new Layer(784, Layer.ActivationFunction.ReLU);
LTEST[1] = new Layer(200, Layer.ActivationFunction.ReLU);
LTEST[2] = new Layer(80, Layer.ActivationFunction.ReLU);
LTEST[3] = new Layer(10, Layer.ActivationFunction.Softmax);
LConnection[] CTEST = new LConnection[3];
CTEST[0] = new LConnection(LTEST[0], LTEST[1]);
CTEST[1] = new LConnection(LTEST[1], LTEST[2]);
CTEST[2] = new LConnection(LTEST[2], LTEST[3]);
NNTEST = new NeuralNetwork(LTEST, CTEST, NeuralNetwork.CostFunctions.CrossEntropy, 0.01);
}
private void Form1_Load(object sender, EventArgs e)
{
Input.Add(new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 1 });
Input.Add(new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 });
Input.Add(new double[] { 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 });
Input.Add(new double[] { 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 });
Input.Add(new double[] { 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 });
Input.Add(new double[] { 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 });
Input.Add(new double[] { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 });
Input.Add(new double[] { 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 });
Input.Add(new double[] { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 });
Input.Add(new double[] { 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 });
//Layer[] L1 = new Layer[2];
//L1[0] = new Layer(784, Layer.ActivationFunction.ReLU);
//L1[1] = new Layer(200, Layer.ActivationFunction.ReLU);
//Layer[] L2 = new Layer[2];
//L2[0] = new Layer(80, Layer.ActivationFunction.ReLU);
//L2[1] = new Layer(10, Layer.ActivationFunction.Softmax);
//LConnection[] C1 = new LConnection[1];
//C1[0] = new LConnection(L1[0], L1[1]);
//LConnection[] C2 = new LConnection[1];
//C2[0] = new LConnection(L2[0], L2[1]);
//NeuralNetwork[] NN = new NeuralNetwork[2];
//NN[0] = new NeuralNetwork(L1, C1,0.01);
//NN[1] = new NeuralNetwork(L2, C2,0.01);
//NNConnection[] NNC = new NNConnection[1];
//NNC[0] = new NNConnection(NN[0], NN[1]);
//M = new Model(NN, NNC, Model.CostFunctions.CrossEntropy,0.01);
D0 = new Dataset(784, 10);
D0.LoadTesting("mnist_test.csv");
D0.LoadTraining("mnist_test.csv");
D1 = new Dataset(10, 10);
D1.TrainingInput = Input;
D1.TrainingLable = Input;
Layer[] L0 = new Layer[2];
L0[0] = new Layer(10, Layer.ActivationFunction.Sigmoid);
L0[1] = new Layer(20, Layer.ActivationFunction.Sigmoid);
Random rand = new Random();
LConnection[] C0 = new LConnection[1];
C0[0] = new LConnection(L0[0], L0[1], rand);
NN[0] = new NeuralNetwork(L0, C0, 0.1);
Layer[] L1 = new Layer[3];
L1[0] = new Layer(784, Layer.ActivationFunction.ReLU);
L1[1] = new Layer(200, Layer.ActivationFunction.ReLU);
L1[2] = new Layer(10, Layer.ActivationFunction.Softmax);
LConnection[] C1 = new LConnection[2];
C1[0] = new LConnection(L1[0], L1[1], rand);
C1[1] = new LConnection(L1[1], L1[2], rand);
NN[1] = new NeuralNetwork(L1, C1, NeuralNetwork.CostFunctions.CrossEntropy, 0.01);
NNConnection[] NNC = new NNConnection[1];
NNC[0] = new NNConnection(NN[0], NN[1]);
M = new Model(NN, NNC, Model.CostFunctions.MeanSquareSrror, 0.1);
}
