private void createNetwork(Perceptron[] ps)
{
if (perceptrons == null)
{
perceptrons = new Perceptron[neurons_per_layer.Sum()];
int p_index = 0;
for (int i = 0; i < hidden_layers_count; i++)
{
// Check if is last layer (output layer)
bool isOutputLayer = false;
if (i == hidden_layers_count - 1)
{
isOutputLayer = true;
}
for (int p = 0; p < neurons_per_layer[i]; p++)
{
int input_nodes;
if (i == 0)
{
input_nodes = this.input_nodes;
}
else
{
input_nodes = neurons_per_layer[i - 1];
}
perceptrons[p_index++] = new Perceptron(input_nodes, random, isOutputLayer);
}
}
}
else
{
perceptrons = ps;
}
}
public double[] Output => Neurons.Select(n => n.Output).ToArray(); //Linq expressions help with basic selections
/// <summary>
/// Constructs a Layer object
/// </summary>
/// <param name="activation">the activation function to be set in all neurons of the constructed layer</param>
/// <param name="inputCount">the number of inputs of the layer</param>
/// <param name="neuronCount">the number of neurons in the layer</param>
public Layer(Func <double, double> activation, int inputCount, int neuronCount)
{
Neurons = new Perceptron[neuronCount];
for (int i = 0; i < Neurons.Length; i++)
{
Neurons[i] = new Perceptron(activation, inputCount);
}
}
private static Perceptron Train_And()
{
var ts = new TrainingSet_And();
var perceptron = new Perceptron(ts);
perceptron.TrainPerceptron(8);
TestLine(0, 0, perceptron.CalculateOutput(0, 0));
TestLine(1, 0, perceptron.CalculateOutput(1, 0));
TestLine(0, 1, perceptron.CalculateOutput(0, 1));
TestLine(1, 1, perceptron.CalculateOutput(1, 1));
return(perceptron);
}
// Split genome to cross-over in genetic controller
public object[] crossOverSplit()
{
int p1_count = (int)Math.Ceiling(perceptrons.Length / 2.0), p2_count = perceptrons.Length - p1_count;
Perceptron[] p1 = new Perceptron[p1_count], p2 = new Perceptron[p2_count];
for (int i = 0; i < p1_count; i++)
{
p1[i] = perceptrons[i];
}
for (int i = p1_count; i < p2_count; i++)
{
p2[i] = perceptrons[i];
}
return(new object[] { p1, p2 });
}
public Neural(int inputNodes, int hiddenNodes, int outputNodes)
{
m_iNumInputNodes = inputNodes;
m_iNumHiddenNodes = hiddenNodes;
m_iNumOutputNodes = outputNodes;
int i, j, k;
m_arInputNodes = new Perceptron[m_iNumInputNodes + 1];
for (i = 0; i <= m_iNumInputNodes; i++)
{
m_arInputNodes[i] = new Perceptron(PerceptionType.PERCEPTION_INPUT, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arInputNodes[m_iNumInputNodes].SetBiasNode();
m_arHiddenNodes = new Perceptron[m_iNumHiddenNodes + 1];
for (j = 0; j <= m_iNumHiddenNodes; j++)
{
m_arHiddenNodes[j] = new Perceptron(PerceptionType.PERCEPTION_HIDDEN, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arHiddenNodes[m_iNumHiddenNodes].SetBiasNode();
m_arOutputNodes = new Perceptron[m_iNumOutputNodes];
for (k = 0; k < m_iNumOutputNodes; k++)
{
m_arOutputNodes[k] = new Perceptron(PerceptionType.PERCEPTION_OUTPUT, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arInputHiddenConn = new double[m_iNumInputNodes + 1, m_iNumHiddenNodes];
m_arHiddenOutputConn = new double[m_iNumHiddenNodes + 1, m_iNumOutputNodes];
Random rand = new Random();
for (i = 0; i <= m_iNumInputNodes; i++)
{
for (j = 0; j < m_iNumHiddenNodes; j++)
{
m_arInputHiddenConn[i, j] = rand.NextDouble();
}
}
for (j = 0; j <= m_iNumHiddenNodes; j++)
{
for (k = 0; k < m_iNumOutputNodes; k++)
{
m_arHiddenOutputConn[j, k] = rand.NextDouble();
}
}
}
public Neural(int inputNodes, int hiddenNodes, int outputNodes)
{
m_iNumInputNodes = inputNodes;
m_iNumHiddenNodes = hiddenNodes;
m_iNumOutputNodes = outputNodes;
int i, j, k;
m_arInputNodes = new Perceptron[m_iNumInputNodes + 1];
for (i = 0; i <= m_iNumInputNodes; i++)
{
m_arInputNodes[i] = new Perceptron(PerceptionType.PERCEPTION_INPUT, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arInputNodes[m_iNumInputNodes].SetBiasNode();
m_arHiddenNodes = new Perceptron[m_iNumHiddenNodes + 1];
for (j = 0; j <= m_iNumHiddenNodes; j++)
{
m_arHiddenNodes[j] = new Perceptron(PerceptionType.PERCEPTION_HIDDEN, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arHiddenNodes[m_iNumHiddenNodes].SetBiasNode();
m_arOutputNodes = new Perceptron[m_iNumOutputNodes];
for (k = 0; k < m_iNumOutputNodes; k++)
{
m_arOutputNodes[k] = new Perceptron(PerceptionType.PERCEPTION_OUTPUT, ActionvationFunction.SIGMOID_FUNCTION);
}
m_arInputHiddenConn = new double[m_iNumInputNodes + 1, m_iNumHiddenNodes];
m_arHiddenOutputConn = new double[m_iNumHiddenNodes + 1, m_iNumOutputNodes];
Random rand = new Random();
for (i = 0; i <= m_iNumInputNodes; i++)
{
for (j = 0; j < m_iNumHiddenNodes; j++)
{
m_arInputHiddenConn[i, j] = rand.NextDouble();
}
}
for (j = 0; j <= m_iNumHiddenNodes; j++)
{
for (k = 0; k < m_iNumOutputNodes; k++)
{
m_arHiddenOutputConn[j, k] = rand.NextDouble();
}
}
}
public Form1()
{
InitializeComponent();
PerceptronInputs = new List <Input>();
myDrawing = new MyDrawing();
// Transfer function 0 == binary, 1 == sigmoid
_neuralNetwork = new NeuralNetwork.NeuralNetwork(2, 2, 1);
_perceptron = new Perceptron(ParseData("AND.txt"));
//_perceptron.PerceptronNeuron.TrainUntil(_perceptron.TrainingSet, 0.2);
myDrawing.DrawPerceptron(_perceptron, perceptronPictureBox);
InitControls();
refreshNetwork();
errorChart.Series.Clear();
//Hopfield
_hopfield = new Hopfield(5, 7, 50);
_hopImage = new List <double>();
}
private void button2_Click(object sender, EventArgs e)
{
if (openFileDialog1.ShowDialog() == DialogResult.OK)
{
string sFileName = openFileDialog1.FileName;
_perceptron = new Perceptron(ParseData(sFileName));
PerceptronInputs.Clear();
for (int i = 0; i < _perceptron.TrainingSet[0].Item1.Length; i++)
{
PerceptronInputs.Add(new Input()
{
Index = (i + 1), Name = "Input " + (i + 1)
});
}
comboInputs.DataSource = PerceptronInputs;
myDrawing.DrawPerceptron(_perceptron, perceptronPictureBox);
}
}
public void DrawPerceptron(Perceptron perceptron, PictureBox picBox)
{
PerceptronHeight = 100;
PerceptronWidth = 100;
var inputCount = perceptron.TrainingSet[0].Item1.Length;
var inputHeight = picBox.Height / inputCount;
var inputSpace = 50;
var pic = new Bitmap(picBox.Height, picBox.Width);
var g = Graphics.FromImage(pic);
var penNeuron = new Pen(Color.Black, 2);
var penInput = new Pen(Color.Black, 1);
var penRect = new Pen(Color.Black, 1);
var brush = Brushes.Orange;
g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
g.TextRenderingHint = System.Drawing.Text.TextRenderingHint.AntiAliasGridFit;
g.SmoothingMode = System.Drawing.Drawing2D.SmoothingMode.AntiAlias;
var centerX = (pic.Width / 2) - (PerceptronWidth / 2);
var centerY = (pic.Height / 2) - (PerceptronHeight / 2);
var picCenterX = (pic.Width / 2);
var picCenterY = (pic.Height / 2);
g.DrawEllipse(penNeuron, centerX, centerY, PerceptronWidth, PerceptronHeight);
var centerPoint = new PointF(picCenterX, picCenterY);
//var connectionPoint = new PointF(PerceptronWidth / 10, (pic.Height / 3));
//var intersectPoint = ClosestIntersection(picCenterX, picCenterY, PerceptronWidth/2, connectionPoint, centerPoint);
//g.DrawLine(pen, connectionPoint, intersectPoint);
//connectionPoint = new PointF(PerceptronWidth / 10, pic.Height / 3);
//intersectPoint = LineIntersectEllipse(centerPoint, connectionPoint, PerceptronHeight, PerceptronWidth);
// X = 0, Y = 461
// X = 0; Y = 661
var linePoints = (inputCount - 1) * inputSpace;
var inputX = 50;
var inputY = picCenterY - (linePoints / 2);
var pointA = new PointF();
var pointB = new PointF();
#region Inputs
for (int i = 0; i < perceptron.TrainingSet.Count; i++)
{
for (int j = 0; j < perceptron.TrainingSet[i].Item1.Length; j++)
{
pointA = new PointF(inputX, inputY + (inputSpace * j));
pointB = ClosestIntersection(picCenterX, picCenterY, PerceptronWidth / 2, pointA, centerPoint);
//var rect = new RectangleF(inputX -50, inputY + (inputSpace * j) - (inputSpace / 4), 50, 25);
//g.DrawString("Input " + (j+1), new Font("Arial", 10, FontStyle.Regular), Brushes.Black, rect);
//g.DrawRectangle(penRect, Rectangle.Round(rect));
g.DrawEllipse(penNeuron, inputX - inputSpace, inputY + (inputSpace * j) - (inputSpace / 2), 50, 50);
g.DrawString("Input " + (j + 1), new Font("Arial", 8, FontStyle.Regular), Brushes.Black, inputX - 40,
inputY + (inputSpace * j) - (inputSpace / 5));
g.DrawLine(penInput, pointA, pointB);
g.DrawString("Weight " + (j + 1) + ": " + Math.Round(perceptron.PerceptronNeuron.InputSynapses[j].Weight, 2), new Font("Arial", 8, FontStyle.Regular), Brushes.Black, inputX + (inputSpace / 5), inputY + (inputSpace * j) + (inputSpace / 5));
//g.DrawString(value.ToString("0.00"), new Font("Arial", 7), Brushes.Green, X + 5, Y + 10);
}
}
#endregion
#region Output
pointA = new PointF(picCenterX + (PerceptronWidth / 2), picCenterY);
pointB = new PointF(picCenterX + (PerceptronWidth / 2) + 75, picCenterY);
g.DrawLine(penInput, pointA, pointB);
g.DrawEllipse(penNeuron, picCenterX + (PerceptronWidth / 2) + 75, picCenterY - (50 / 2), 50, 50);
g.DrawString("Out: " + perceptron.PerceptronNeuron.Output, new Font("Arial", 8, FontStyle.Regular), Brushes.Black, picCenterX + (PerceptronWidth / 2) + 85,
picCenterY - (50 / 6));
#endregion
//var pointA = new PointF(inputX, inputY);
//var pointB = new PointF(inputX, inputY + inputSpace);
//g.DrawLine(pen, pointA, pointB);
//pointA = new PointF(25, picCenterY - (PerceptronWidth / 2));
//pointB = new PointF(25, picCenterY + (PerceptronWidth / 2));
//g.DrawLine(pen, pointA, pointB);
picBox.Image = pic;
}
void setup()
{
perceptron = new Perceptron(3);
// generate trainers
Random random = new Random();
for (int i = 0; i < trainers.Count(); i++)
{
float x = (float)((this.Width * random.NextDouble()) - (this.Width / 2));
float y = (float)((this.Height * random.NextDouble()) - (this.Height / 2));
int answer = 1;
if (y < lineFunction(x))
answer = -1;
trainers[i] = new Trainer(x, y, answer);
}
}
