static void Main(string[] args)
{
NeuralNetwork nn = new NeuralNetwork();
nn.AddLayer(2, new Sigmoid());
nn.AddLayer(3, new Sigmoid());
nn.AddLayer(3, new Sigmoid());
nn.AddLayer(1, new Sigmoid());
nn.Build();
double[][] input = new double[][] {
new double[] { 0, 0 },
new double[] { 0, 1 },
new double[] { 1, 0 },
new double[] { 1, 1 }
};
double[][] and = new double[][] {
new double[] { 0 },
new double[] { 0 },
new double[] { 0 },
new double[] { 1 },
};
double[][] or = new double[][] {
new double[] { 0 },
new double[] { 1 },
new double[] { 1 },
new double[] { 1 },
};
double[][] xor = new double[][] {
new double[] { 0 },
new double[] { 1 },
new double[] { 1 },
new double[] { 0 },
};
nn.Train(input, xor, 10000, 0.25);
nn.Predict(input[0]);
nn.Predict(input[1]);
nn.Predict(input[2]);
nn.Predict(input[3]);
Console.ReadLine();
}
private static void Main(string[] args)
{
Console.WriteLine("Loading samples from files...");
var inputs = LoadFromCsv("X.csv").ToArray();
var outputs = LoadFromCsv("y.csv").Select(ConvertOutput).ToArray();
if (inputs.Length == 0)
{
return;
}
var net = new NeuralNetwork(new SigmoidFunction());
net.AddLayer(400);
net.AddLayer(25);
net.AddLayer(10);
const int numEpochs = 100;
const int batchSize = 50;
const double learningSpeed = 0.21;
int numSamples = inputs.Length;
int numBatches = numSamples / batchSize;
Console.WriteLine($"Started learning network ({numEpochs} epochs, {numBatches} minibatches, {batchSize} samples each)");
var sw = new Stopwatch();
for (int i = 0; i < numEpochs; ++i)
{
sw.Start();
var shuffledBatch = inputs
.Zip(outputs, Tuple.Create)
.OrderBy(x => Guid.NewGuid())
.ToArray();
var shuffledInputs = shuffledBatch
.Select(x => x.Item1)
.ToArray();
var shuffledOutputs = shuffledBatch
.Select(x => x.Item2)
.ToArray();
// Perform Stochastic Gradient Descent
for (int j = 0; j < numBatches; ++j)
{
var batchInputs = shuffledInputs
.Skip(j * batchSize)
.Take(batchSize)
.ToArray();
var batchOutputs = shuffledOutputs
.Skip(j * batchSize)
.Take(batchSize)
.ToArray();
net.UpdateMultipleSamples(batchInputs, batchOutputs, learningSpeed);
Console.Write('.');
}
Console.WriteLine();
int numCorrect = inputs
.Select(input => net.GetOutput(input))
.Where((output, j) => outputs[j][output.MaximumIndex()].Equals(1.0))
.Count();
int numInvalid = numSamples - numCorrect;
double ratio = (double)numInvalid / numSamples;
Console.WriteLine($"Epoch: {i + 1} / {numEpochs}; Error: {numInvalid} / {numSamples} [{ratio}]; Time: {sw.Elapsed}");
sw.Reset();
}
}
