如何从Accord.net SOM（Kohonen's）中的网络进行类/标签分配

Question

任何人都可以告诉如何从经过培训的网络中获取/分配集群类/标签。

这是代码示例，可让您了解如何执行它：

Accord.Math.Random.Generator.Seed = 0;

int numberOfInputs = 3;
int hiddenNeurons = 25;

double[][] input =
{
     new double[] { -1, -1, -1 },
     new double[] { -1,  1, -1 },
     new double[] {  1, -1, -1 },
     new double[] {  1,  1, -1 },
     new double[] { -1, -1,  1 },
     new double[] { -1,  1,  1 },
     new double[] {  1, -1,  1 },
     new double[] {  1,  1,  1 },
     // ...
};

var network = new DistanceNetwork(numberOfInputs, hiddenNeurons);
var teacher = new SOMLearning(network);
double error = double.PositiveInfinity;

for (int i = 0; i < 1000; i++)
    error = teacher.RunEpoch(input);

// how can I know/assign class/label of each item in input array?

Answer 1

Accord示例可能会提供进一步的帮助：

群集（SOM）

特别是本节：

for (int y = 0, i = 0; y < 100; y++)
{
    // for all pixels
    for (int x = 0; x < 100; x++, i++, ptr += 6)
    {
        Neuron neuron = layer.Neurons[i];

        // red
        ptr[2] = ptr[2 + 3] = ptr[2 + stride] = ptr[2 + 3 + stride] =
            (byte)Math.Max(0, Math.Min(255, neuron.Weights[0]));

        // green
        ptr[1] = ptr[1 + 3] = ptr[1 + stride] = ptr[1 + 3 + stride] =
            (byte)Math.Max(0, Math.Min(255, neuron.Weights[1]));

        // blue
        ptr[0] = ptr[0 + 3] = ptr[0 + stride] = ptr[0 + 3 + stride] =
            (byte)Math.Max(0, Math.Min(255, neuron.Weights[2]));
    }

    ptr += offset;
    ptr += stride;
}