C# - 3층 구조의 신경망

아래의 책에 보면,

실체가 손에 잡히는 딥러닝
; http://www.yes24.com/Product/Goods/74258238

# http://nanya-kanya.net/index.php/1232/#outline__1_29

import numpy as np
import matplotlib.pyplot as plt

X = np.arange(-1.0, 1.0, 0.2)
Y = np.arange(-1.0, 1.0, 0.2)

Z = np.zeros((10, 10))

w_im = np.array([[2.0, -2.0],
                 [1.0, 4.0]])
w_mo = np.array([[1.0],
                 [-1.0]])

b_im = np.array([3.0, -3.0])
b_mo = np.array([0.1])

def middle_layer(x, w, b):
    u = np.dot(x, w) + b
    return 1 / (1 + np.exp(-u))

def output_layer(x, w, b):
    u = np.dot(x, w) + b
    return u

for i in range(10):
    for j in range(10):

        inp = np.array([X[i], Y[j]])
        mid = middle_layer(inp, w_im, b_im)
        out = output_layer(mid, w_mo, b_mo)

        Z[j][i] = out[0]
        
plt.imshow(Z, "gray", vmin = 0.0, vmax = 1.0)
plt.colorbar()
plt.show()

using MathNet.Numerics.LinearAlgebra;
using System;
using System.Linq;

using np = PythonUtils;
using vector = MathNet.Numerics.LinearAlgebra.Vector<double>;
using matrix = MathNet.Numerics.LinearAlgebra.Matrix<double>;

class Program
{
    static void Main(string[] args)
    {
        var X0 = np.arange(-1.0, 1.0, 0.2).ToArray();
        var X1 = np.arange(-1.0, 1.0, 0.2).ToArray();

        double[,] Y = new double[X0.Length, X1.Length];

        matrix w_im = GetMatrix(new[] { -4.0, 4.0 }, new[] { -4.0, -4.0 });
        matrix w_mo = GetMatrix(new[] { 1.0 }, new[] { -1.0 });

        vector b_im = GetVector(3.0, -3.0);
        vector b_mo = GetVector(0.1);

        Func<vector, matrix, vector, vector> middle_layer = (x, w, b) =>
        {
            vector u = x * w + b;
            return 1 / (1 + np.exp(-u));
        };

        Func<vector, matrix, vector, vector> output_layer = (x, w, b) =>
        {
            return x * w + b;
        };

        for (int i = 0; i < X0.Length; i++)
        {
            for (int j = 0; j < X1.Length; j++)
            {
                var inp = GetVector(X0[i], X1[j]);
                var mid = middle_layer(inp, w_im, b_im);
                var outp = output_layer(mid, w_mo, b_mo);

                Y[j, i] = outp[0];
            }
        }

        OutputImage("layer3_neuron.png");

        void OutputImage(string fileName)
        {
            Gridmap grid = new Gridmap(371, 371);
            grid.Show(Y, fileName);
        }
    }

    private static Matrix<double> GetMatrix(params double[][] values)
    {
        return CreateMatrix.DenseOfRows(values.Length, values[0].Length, values);
    }

    private static Vector<double> GetVector(params double [] values)
    {
        return CreateVector.DenseOfArray(values);
    }
}

double minX = gridmap.Min();
double maxX = gridmap.Max();

Func<double, double> lerf = (value) =>
{
    return (value - minX) / (maxX - minX);
};

for (int i = 0; i < count; i++)
{
    double h, l, s;
    double r1, g1, b1;

    h = 0;
    l = lerf(gridmap[i]);
    s = 0;

    pl.hlsrgb(h, l, s, out r1, out g1, out b1);
    r[i + 16] = (int)(r1 * 255.0);
    g[i + 16] = (int)(g1 * 255.0);
    b[i + 16] = (int)(b1 * 255.0);
}