private void button2_Click(object sender,EventArgs e)
{
int cycles=int.Parse(textBox1.Text);
int hiddenNeurons=int.Parse(textBox2.Text);
double learnRate=double.Parse(textBox3.Text);
double momentum=double.Parse(textBox4.Text);
int[] layers = new int[] { 1,hiddenNeurons,1 };
MultilayerPerceptron ann=new MLPGenerator().Create(layers,1,new Sigmoid(2));
ann.Reset(-1,1);
ann.Momentum=momentum;
ann.LearnFactor=learnRate;
List<TrainingData> trainingList=new List<TrainingData>();
for(int i=0;i<Map.Count;i++) {
double[] input=new double[] { (double)Map[i].Item1/(double)pictureBox1.Width };
double[] output=new double[] { (double)Map[i].Item2/(double)pictureBox1.Height };
trainingList.Add(new TrainingData(input,output));
}
var res=ann.BP(new BPRequest(trainingList.ToArray(),cycles));
for(int i=0;i<pictureBox1.Width;i++) {
int x = i;
double yd = ann.Pulse(new double[] { (double)i/(double)pictureBox1.Width })[0];
int y=Convert.ToInt32(yd*pictureBox1.Height);
Point p= new Point(x,y);
Graph.DrawRectangle(new Pen(new SolidBrush(Color.Orange)),p.X,p.Y,1,1);
}
pictureBox1.Image=Bitmap;
}
public void ResetAutoTest()
{
int[] layers = new int[] { 4,5,1 };
MultilayerPerceptron nn = new MLPGenerator().Create(layers,1,new Sigmoid());
double[] randomSelected=new double[] {
nn.Structure.Input[1].Next[1].Weight.Value,
nn.Structure.Input[1].Next[2].Weight.Value,
nn.Structure.Input[1].Next[2].Next.Next[0].Weight.Value,
nn.Structure.Output[0].Previous[1].Weight.Value
};
double[] inp = new double[] { -0.978, 2.34, 0.2, -0.33};
double[] pulsed=nn.Pulse(inp);
nn.Reset(0,1);
double[] randomSelected2=new double[] {
nn.Structure.Input[1].Next[1].Weight.Value,
nn.Structure.Input[1].Next[2].Weight.Value,
nn.Structure.Input[1].Next[2].Next.Next[0].Weight.Value,
nn.Structure.Output[0].Previous[1].Weight.Value
};
for(int i=0;i<randomSelected.Length;i++) {
Assert.AreNotEqual(randomSelected[i],randomSelected2[i]);
}
double[] pulsed2=nn.Pulse(inp);
for(int i=0;i<pulsed.Length;i++) {
Assert.AreNotEqual(pulsed[i],pulsed2[i]);
}
}