private void btnTrainOne_Click(object sender, EventArgs e)
{
if (net == null)
{
return;
}
Sample fig = generator.GenerateFigure();
pictureBox1.Image = generator.genBitmap();
pictureBox1.Invalidate();
net.Train(fig);
set_result(fig);
}
static void Main(string[] args)
{
//Simple Artificial Neural Network
//We will construct a 3 layer ANN to learn the XOR function.
//Define the network layers
//layer 0 is the input layer, 1 is the hidden layer, 2 is the output layer.
int[] layers = new[] { 2, 2, 1 };
var nn = new NeuralNetwork(layers)
{
Iterations = 1000, //training iterations
Alpha = 3.5, //learning rate, lower is slower, too high may not converge.
L2_Regularization = true, //set L2 regularization to prevent overfitting
Lambda = 0.0003, //strength of L2
Rnd = new Random(12345) //provide a seed for repeatable outputs
};
//Define the inputs and the outputs.
//The last value in these training sets is the expected output.
//Simplest, learn that the first input value is always the answer
//var training = new double[][]
//{
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 1, 0 },
// new double[]{ 1, 0, 1 },
// new double[]{ 0, 1, 0 },
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 0, 0 },
//};
//Simple AND
//var training = new double[][]
//{
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 1, 0 },
// new double[]{ 1, 0, 0 },
// new double[]{ 0, 1, 0 },
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 0, 0 },
//};
//Simple OR
//var training = new double[][]
//{
// new double[]{ 1, 0, 1 },
// new double[]{ 0, 1, 1 },
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 0, 0 },
// new double[]{ 1, 1, 1 },
// new double[]{ 0, 0, 0 },
// new double[]{ 1, 0, 1 },
// new double[]{ 0, 1, 1 },
//};
//Simple XOR
var training = new double[][]
{
new double[] { 1, 0, 1 },
new double[] { 0, 1, 1 },
new double[] { 1, 1, 0 },
new double[] { 0, 0, 0 },
new double[] { 1, 1, 0 },
new double[] { 0, 0, 0 },
new double[] { 1, 0, 1 },
new double[] { 0, 1, 1 },
};
//Normalize the input to -1,+1 ?
//We want a 0 mean, and 1 stdev.
//Take the first 2 columns as input, and last 1 column as target y (the expected label)
var input = new double[training.GetLength(0)][];
for (int i = 0; i < training.GetLength(0); i++)
{
input[i] = new double[layers[0]];
for (int j = 0; j < layers[0]; j++)
{
input[i][j] = training[i][j];
}
}
//Create the expected label array
var y = new double[training.GetLength(0)];
for (int i = 0; i < training.GetLength(0); i++)
{
y[i] = training[i][layers[0]];
}
//Let's also monitor training by providing a delegate function
nn.Monitor = delegate(TrainingTelemetry t)
{
Console.CursorLeft = 0;
Console.CursorTop = 0;
//Display some information about its learning at each iteration
Console.WriteLine($"Iteration {t.Iteration}");
//Display some sample data
Console.WriteLine($"{nn.Predict(new[] { 0.0, 0.0 })[0]} -> 0");
Console.WriteLine($"{nn.Predict(new[] { 0.0, 1.0 })[0]} -> 1");
Console.WriteLine($"{nn.Predict(new[] { 1.0, 0.0 })[0]} -> 1");
Console.WriteLine($"{nn.Predict(new[] { 1.0, 1.0 })[0]} -> 0");
//Just for fun lets print out the weights and biases
Console.WriteLine("\nWeights:");
for (int l = 0; l < t.Weights.Length; l++)
{
Console.WriteLine($" Layer {l}");
Console.WriteLine(" --------------------------");
for (int j = 0; j < t.Weights[l].GetLength(0); j++)
{
for (int k = 0; k < t.Weights[l].GetLength(1); k++)
{
Console.Write(" {0:#.##}\t", t.Weights[l][j, k]);
}
Console.WriteLine();
}
}
Console.WriteLine("\nBiases:");
Console.WriteLine("--------------------------");
for (int l = 1; l < t.Bias.Length; l++)
{
for (int n = 0; n < t.Bias[l].Length; n++)
{
Console.Write(" {0:#.##}\t", t.Bias[l][n]);
}
Console.WriteLine();
}
//Display average error
var absCost = (double)t.Error.Sum(v => Math.Abs(v)) / t.Error.Length;
Console.WriteLine("\nError {0:#.#####}", absCost);
//You can step through manually if you wish
//Console.ReadKey(true);
};
//Begin training the network to learn the function that matches our data.
nn.Train(input, y);
//Confirm that its worked
Console.WriteLine($"The network learned XOR(1,0)={nn.Predict(new[] { 1.0, 0.0 })[0]}");
Console.WriteLine($"The network learned XOR(1,1)={nn.Predict(new[] { 1.0, 1.0 })[0]}");
Console.WriteLine("press any key to continue");
Console.ReadKey(true);
}