List<Neuron> getNeighborhoodNeurons(Tuple<int, int> neighborhoodTemplate, Tuple<int, int> position, Neuron[,] network)
{
List<Neuron> neighborhoodNeurons = new List<Neuron>();
for (int i = 1; i <= neighborhoodTemplate.Item1; i++)
{
for (int j = 1; j <= neighborhoodTemplate.Item2; j++)
{
if (!(position.Item2 - j < 0))
{
if (!(position.Item1 - i < 0))
{
neighborhoodNeurons.Add(network[position.Item1 - i, position.Item2 - j]);
}
if (!(position.Item1 + i >= network.GetLength(0) - 1))
{
neighborhoodNeurons.Add(network[position.Item1 + i, position.Item2 - j]);
}
}
if (!(position.Item2 + j >= network.GetLength(1) - 1))
{
if (!(position.Item1 - i < 0))
{
neighborhoodNeurons.Add(network[position.Item1 - i, position.Item2 + j]);
}
if (!(position.Item1 + i >= network.GetLength(0) - 1))
{
neighborhoodNeurons.Add(network[position.Item1 + i, position.Item2 + j]);
}
}
}
if (!(position.Item1 + i >= network.GetLength(0) - 1))
{
neighborhoodNeurons.Add(network[position.Item1 + i, position.Item2]);
}
if (!(position.Item1 - i < 0))
{
neighborhoodNeurons.Add(network[position.Item1 - i, position.Item2]);
}
}
for (int j = 1; j <= neighborhoodTemplate.Item2; j++)
{
if (!(position.Item2 - j < 0))
{
neighborhoodNeurons.Add(network[position.Item1, position.Item2 - j]);
}
if (!(position.Item2 + j >= network.GetLength(1) - 1))
{
neighborhoodNeurons.Add(network[position.Item1, position.Item2 + j]);
}
}
return neighborhoodNeurons;
}
public void processOutput(Tuple<int, int> neighborhoodTemplate, Tuple<int, int> position, Neuron[,] network)
{
double output = 0;
List<Neuron> neighborhoodNeurons = getNeighborhoodNeurons(neighborhoodTemplate, position, network);
output += this.input * weight[0];
output += this.bias * weight[1];
output += processATemplate(neighborhoodNeurons) * weight[2];
output += processBTemplate(neighborhoodNeurons) * weight[3];
output = 1 / (1 + Math.Exp(-output));
this.output = output;
}
public void processOutputTest()
{
Neuron[,] testNeurons = new Neuron[5,5];
double expected;
double actual;
for (int i = 0; i < testNeurons.GetLength(0); i++)
{
for (int j = 0; j < testNeurons.GetLength(1); j++)
{
testNeurons[i, j] = new Neuron(1);
testNeurons[i, j].setInput(1);
testNeurons[i, j].setWeight(new double[] { 1, 1, 1, 1 });
}
}
Neuron testedNeuron = testNeurons[2, 2];
testedNeuron.processOutput(new Tuple<int, int>(1, 1), new Tuple<int, int>(2, 2), testNeurons);
expected = 1 / (1 + Math.Exp(-10));
actual = testedNeuron.getOutput();
Assert.AreEqual(expected, actual);
testedNeuron = testNeurons[0, 0];
testedNeuron.processOutput(new Tuple<int, int>(1, 1), new Tuple<int, int>(0, 0), testNeurons);
expected = 1 / (1 + Math.Exp(-5));
actual = testedNeuron.getOutput();
Assert.AreEqual(expected, actual);
testedNeuron = testNeurons[4, 4];
testedNeuron.processOutput(new Tuple<int, int>(1, 1), new Tuple<int, int>(4, 4), testNeurons);
expected = 1 / (1 + Math.Exp(-5));
actual = testedNeuron.getOutput();
Assert.AreEqual(expected, actual);
testedNeuron = testNeurons[0, 4];
testedNeuron.processOutput(new Tuple<int, int>(1, 1), new Tuple<int, int>(0, 4), testNeurons);
expected = 1 / (1 + Math.Exp(-5));
actual = testedNeuron.getOutput();
Assert.AreEqual(expected, actual);
testedNeuron = testNeurons[4, 0];
testedNeuron.processOutput(new Tuple<int, int>(1, 1), new Tuple<int, int>(4, 0), testNeurons);
expected = 1 / (1 + Math.Exp(-5));
actual = testedNeuron.getOutput();
Assert.AreEqual(expected, actual);
}
public NeuralNetwork(int dimensionX, int dimensionY, int templateDimensionX, int templateDimensionY, Bitmap inputImage, double bias)
{
this.network = new Neuron[dimensionX, dimensionY];
this.neighborhoodTemplate = new Tuple<int, int>(templateDimensionX, templateDimensionY);
for (int i = 0; i < dimensionX; i++)
{
for (int j = 0; j < dimensionY; j++)
{
network[i, j] = new Neuron(bias);
Color clr = inputImage.GetPixel(i, j);
if (clr.R + clr.G + clr.B == 0)
{
network[i, j].setInput(1);
}
else
{
network[i, j].setInput(0);
}
}
}
}