public static bool Train(NeuralNetwork nn, double[] input, double[] output)
{
if (input.Length != nn.Layers[0].Neurons.Count ||
output.Length != nn.Layers[nn.Layers.Count - 1].Neurons.Count)
{
return(false);
}
nn.Run(input);
for (var i = 0; i < nn.Layers[nn.Layers.Count - 1].Neurons.Count; i++)
{
var neuron = nn.Layers[nn.Layers.Count - 1].Neurons[i];
neuron.Delta = neuron.Value * (1 - neuron.Value) * (output[i] - neuron.Value);
for (var j = nn.Layers.Count - 2; j >= 1; j--)
{
for (var k = 0; k < nn.Layers[j].Neurons.Count; k++)
{
var n = nn.Layers[j].Neurons[k];
n.Delta = n.Value *
(1 - n.Value) *
nn.Layers[j + 1].Neurons[i].Dendrites[k].Weight *
nn.Layers[j + 1].Neurons[i].Delta;
}
}
}
for (var i = nn.Layers.Count - 1; i >= 1; i--)
{
for (var j = 0; j < nn.Layers[i].Neurons.Count; j++)
{
var n = nn.Layers[i].Neurons[j];
n.Bias = n.Bias + nn.LearningRate * n.Delta;
for (var k = 0; k < n.Dendrites.Count; k++)
{
n.Dendrites[k].Weight = n.Dendrites[k].Weight +
nn.LearningRate * nn.Layers[i - 1].Neurons[k].Value * n.Delta;
}
}
}
return(true);
}
private void buttonRun_Click(object sender, EventArgs e)
{
_neuralNetwork.Run(_param.InitValues.ToArray());
}