static void BigTest()
{
LearningSequence<ISupervisedLearningVector> band = new LearningSequence<ISupervisedLearningVector>();
band.Add(new SupervisedLearningVector(new double[2] { -1, -1 }, -1));
band.Add(new SupervisedLearningVector(new double[2] { 1, -1 }, -1));
band.Add(new SupervisedLearningVector(new double[2] { -1, 1 }, -1));
band.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
LearningSequence<ISupervisedLearningVector> bor = new LearningSequence<ISupervisedLearningVector>();
bor.Add(new SupervisedLearningVector(new double[2] { -1, -1 }, -1));
bor.Add(new SupervisedLearningVector(new double[2] { 1, -1 }, 1));
bor.Add(new SupervisedLearningVector(new double[2] { -1, 1 }, 1));
bor.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
LearningSequence<ISupervisedLearningVector> uand = new LearningSequence<ISupervisedLearningVector>();
uand.Add(new SupervisedLearningVector(new double[2] { 0, 0 }, 0));
uand.Add(new SupervisedLearningVector(new double[2] { 1, 0 }, 0));
uand.Add(new SupervisedLearningVector(new double[2] { 0, 1 }, 0));
uand.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
LearningSequence<ISupervisedLearningVector> uor = new LearningSequence<ISupervisedLearningVector>();
uor.Add(new SupervisedLearningVector(new double[2] { 0, 0 }, 0));
uor.Add(new SupervisedLearningVector(new double[2] { 1, 0 }, 1));
uor.Add(new SupervisedLearningVector(new double[2] { 0, 1 }, 1));
uor.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
sb.AppendLine("%Dla matlaba :)");
sb.AppendLine("% Perceptron prosty!");
sb.AppendLine("%Zaleznosc efektywnosci od wag");
sb.AppendLine("%Bipolarny - AND");
int K = 1000;
double ALPHA = 0.8;
double WEIGHT = 0.1;
PerceptronLearningMethod plm = new PerceptronLearningMethod();
plm.Alpha = ALPHA;
double[] wagi = new double[] { 5, 4,3,2, 1.0, 0.9, 0.8, 0.7, 0.6, 0.5, .4, .3,.2,.1,.0};
sb.Append("weights_avg_noepoch_test_uniand_alpha02_k"+K+" = [");
for (int i = 0; i < wagi.Length; i++)
{
sb.Append(wagi[i] + " "
+ testLearn(plm, uand, Neuron.UnipolarActivationFunction, K, wagi[i])
+ ";");
}
sb.AppendLine();
sb.Append("weights_avg_noepoch_test_biand_alpha02_k"+K+" = [");
for (int i = 0; i < wagi.Length; i++)
{
sb.Append(wagi[i] + " "
+ testLearn(plm, band, Neuron.BipolarActivationFunction, K, wagi[i])
+ ";");
}
sb.AppendLine();
sb.AppendLine("%Zaleznosc efektywnosci od alpha");
double[] alfy = new double[] { 64,32, 16,8,4,2,1,.9,.8,.7,.6,.5, .4, .3, .2, .1, .09, .08, .07, .06, .05, .04, .03, .02, .01, .001 };
sb.Append("alpha_avg_noepoch_test_uniand_k"+K+"=[");
for (int i = 0; i < alfy.Length; i++)
{
plm.Alpha = alfy[i];
sb.Append(alfy[i] + " "
+ testLearn(plm, uand, Neuron.UnipolarActivationFunction, K, WEIGHT)
+ ";");
}
sb.AppendLine();
sb.Append("alpha_avg_noepoch_test_band_k"+K+"=[");
for (int i = 0; i < alfy.Length; i++)
{
plm.Alpha = alfy[i];
sb.Append(alfy[i] + " "
+ testLearn(plm, band, Neuron.BipolarActivationFunction, K, WEIGHT)
+ ";");
}
sb.AppendLine();
Console.WriteLine(sb.ToString());
/*
LearningSequence<ISupervisedLearningVector> klocki = new LearningSequence<ISupervisedLearningVector>();
// kolumny seria 1 zamalowane, seria 2 niezamalowane
// kolejno kwadraty i kola -> duzy, maly
double[][] aa = new double[20][]
// zamalowane | niezamalowane
{ /// dk mk do mo dk mk do mo
new double[] { 1, -1, 1, 1, 1, 1, 1, 1 },
new double[] { -1, -1, 1, 1, 1, -1, -1, 1 },
new double[] { 1, 1, 1, 1, 1, 1, 1, 1 },
new double[] { 1, -1, 1, -1, 1, -1, 1, -1 },
new double[] { 1, 1, 1, 1, 1, 1, 1, 1 },
new double[] { -1, -1, 1, 1, 1, -1, -1, 1 },
new double[] { -1, 1, 1, -1, 1, -1, 1, -1 },
new double[] { 1, 1, 1, 1, 1, 1, 1, -1 },
new double[] { -1, -1, 1, -1, 1, 1, -1, -1 },
new double[] { -1, 1, 1, 1, 1, -1, -1, 1 },
new double[] { -1, 1, 1, -1, -1, 1, 1, -1 },
new double[] { -1, -1, 1, -1, -1, 1, -1, 1 },
new double[] { 1, 1, -1, -1, 1, 1, -1, 1 },
new double[] { 1, 1, -1, 1, 1, 1, -1, 1 },
new double[] { -1, -1, 1, -1, -1, -1, 1, 1 },
new double[] { 1, -1, -1, 1, 1, -1, 1, -1 },
new double[] { 1, 1, -1, -1, -1, -1, -1, -1 },
new double[] { 1, 1, -1, -1, 1, 1, 1, -1 },
new double[] { -1, -1, -1, -1, 1, 1, -1, -1 },
new double[] { 1, -1, -1, 1, -1, -1, 1, -1 }
};
double[][] bb = new double[20][];
for (int i = 0; i < 20; i++)
{
bb[i] = new double[8];
for (int j = 0; j < 8; j++)
{
bb[i][j] = aa[i][j] < 0 ? 0 : 1;
}
}
LearningSequence<ISupervisedLearningVector> klockibb = new LearningSequence<ISupervisedLearningVector>();
for (int i = 0; i < aa.Length; i++)
{
klockibb.Add(new SupervisedLearningVector(bb[i], (i>9?0:1)));
klocki.Add(new SupervisedLearningVector(aa[i], (i>9?-1:1)));
}
NeuralNetwork klockineuron = new NeuralNetwork();
klockineuron.Build(8, null, 1, Neuron.UnipolarActivationFunction, 0.1);
//PerceptronLearningMethod klockilm = new PerceptronLearningMethod();
AdalineLMSLearningMethod klockilm = new AdalineLMSLearningMethod();
klockilm.Epsilon = 1.97;
klockilm.Alpha = 0.01;
klockilm.EpochLimit = 1000;
for (int z = 0; z < 200; z++)
{
klockilm.Learn(klockineuron, klockibb);
if (!klockilm.Failed) Console.WriteLine("Nauczyil sie!");
}
Console.WriteLine("Klocki sie nauczyly");
/*
klockineuron.SetInput(new double[8] { -1, -1, 1, -1, 1, -1, -1, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");
klockineuron.SetInput(new double[8] { 1, -1, 1, -1, -1, 1, -1, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");
klockineuron.SetInput(new double[8] { 1, 1, -1, 1, 1, 1, 1, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");
Console.WriteLine(klockilm.LearningStory.Count + " epok, fail?" + klockilm.Failed);
klockineuron.SetInput(new double[8] { 0, 0, 1, 0, 1, 0, 0, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");
klockineuron.SetInput(new double[8] { 1, 0, 1, 0, 0, 1, 0, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");
klockineuron.SetInput(new double[8] { 1, 1, 0, 1, 1, 1, 1, 1 });
klockineuron.Compute();
Console.WriteLine(klockineuron[0] > 0 ? "lubi " : "nie lubi");*/
}
static double testLearn(ISupervisedLearningMethod lm, LearningSequence<ISupervisedLearningVector> seq, ActivationFunction af, int no, double weight)
{
NeuralNetwork net;
double a = 0;
for (int i = 0; i < no; i++)
{
net = new NeuralNetwork();
net.Build(2, null, 1, af, weight);
lm.Learn(net, seq);
a += lm.LearningStory.Count;
}
return a / no;
}
static void Main(string[] args)
{
//BigTest();
//Console.ReadKey();
//return;
Form1 f = new Form1();
f.ShowDialog();
NeuralNetwork nn = new NeuralNetwork();
nn.Build(2, null, 1, Neuron.BipolarActivationFunction, 0.7);
LearningSequence<ISupervisedLearningVector> seq = new LearningSequence<ISupervisedLearningVector>();
seq.Add(new SupervisedLearningVector(new double[2] { -1, -1 }, -1));
seq.Add(new SupervisedLearningVector(new double[2] { -1, 1 }, -1));
seq.Add(new SupervisedLearningVector(new double[2] { 1, -1 }, -1));
seq.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
LearningSequence<ISupervisedLearningVector> seq2 = new LearningSequence<ISupervisedLearningVector>();
seq2.Add(new SupervisedLearningVector(new double[2] { 0, 0 }, 0));
seq2.Add(new SupervisedLearningVector(new double[2] { 0, 1 }, 0));
seq2.Add(new SupervisedLearningVector(new double[2] { 1, 0 }, 0));
seq2.Add(new SupervisedLearningVector(new double[2] { 1, 1 }, 1));
PerceptronLearningMethod plm = new PerceptronLearningMethod();
AdalineLMSLearningMethod lms = new AdalineLMSLearningMethod();
lms.Alpha = 0.02;
plm.Alpha = 0.2;
lms.Learn(nn, (ILearningSequence<ISupervisedLearningVector>)seq);
nn.SetInput(new double[2] { 0.98, 0.99 });
nn.Compute();
Console.WriteLine(nn[0]);
Console.ReadKey();
}