private void button1_Click(object sender, EventArgs e)
{
Bitmap bitmap = new Bitmap(@"data\\mnist\\2_58.bmp");
this.pictureBox1.Image = bitmap;
BoltzmannPattern pattern = new BoltzmannPattern();
pattern.InputNeurons = bitmap.Width * bitmap.Height;
BasicNetwork network = pattern.Generate();
BoltzmannLogic logic = (BoltzmannLogic)network.Logic;
//CreateCities();
CalculateWeights(network);
}
public void CalculateWeights(BasicNetwork network)
{
int n1, n2, n3, n4;
int i, j;
int predN3, succN3;
double weight;
BoltzmannLogic logic = (BoltzmannLogic)network.Logic;
for (n1 = 0; n1 < _bitmap.Height; n1++)
{
for (n2 = 0; n2 < _bitmap.Width; n2++)
{
i = n1 * _bitmap.Height + n2;
for (n3 = 0; n3 < _bitmap.Height; n3++)
{
for (n4 = 0; n4 < _bitmap.Width; n4++)
{
j = n3 * _bitmap.Height + n4;
weight = 0;
if (i != j)
{
predN3 = (n3 == 0 ? _bitmap.Width - 1 : n3 - 1);
succN3 = (n3 == _bitmap.Width - 1 ? 0 : n3 + 1);
if ((n1 == n3) || (n2 == n4))
{
weight = -gamma;
}
else if ((n1 == predN3) || (n1 == succN3))
{
weight = -distance[n2][n4];
}
}
logic.ThermalSynapse.WeightMatrix[i, j] = weight;
}
}
logic.ThermalLayer.BiasWeights[i] = (-gamma / 2);
}
}
}
