void K_Means(int K)
{
#region Create Output layer
for (int i = 0; i < 5; i++) // Construct the output Layer
{
Perceptron per = new Perceptron(K);
OutPut.Add(per);
}
#endregion
#region Get_Cntres
// Get Random Centers For Clusters
Random rd = new Random();
for (int i = 0; i < K; i++)
{
cluster temp = new cluster();
temp.center = Training_data[rd.Next(0, Training_data.Count)];
Clusters.Add((temp));
sample s = new sample();
s.copysample(temp.center);
old_centers.Add(s);
}
#endregion
old_centers[0].features[0] = (float)-0.0002; // To change it for first time
while (check() )
{
#region updateing_old_centers
for (int i = 0; i < K; i++)
{
Clusters[i].cluster_samples.Clear();
old_centers[i].copysample(Clusters[i].center);
}
#endregion
#region Clustering
List<double> _out = new List<double>();
for (int sampleindx = 0; sampleindx < Training_data.Count(); sampleindx ++ )
{
_out.Clear();
for (int Clusterindx = 0; Clusterindx < Clusters.Count(); Clusterindx ++ )
{
_out.Add(distance(Training_data[sampleindx].features, Clusters[Clusterindx].center.features));
}
Clusters[get_min(_out)].cluster_samples.Add(Training_data[sampleindx]);
}
#endregion
#region Get New Centers
for (int clusterindx = 0; clusterindx < Clusters.Count(); clusterindx ++ )
{
double[] x = new double[Clusters[clusterindx].center.features.Count()];
for (int sampleindx = 0 ; sampleindx < Clusters[clusterindx].cluster_samples.Count ; sampleindx++)
{
for (int featureindx = 0; featureindx < Clusters[clusterindx].center.features.Count(); featureindx++ )
{
x[featureindx] += Clusters[clusterindx].cluster_samples[sampleindx].features[featureindx];
}
}
for (int featureindx = 0; featureindx < Clusters[clusterindx].center.features.Count(); featureindx++)
{
x[featureindx] /= Clusters[clusterindx].cluster_samples.Count();
Clusters[clusterindx].center.features[featureindx] = (float) x[featureindx]; // update new centers
}
}
#endregion
}
}
private void kMeans(int k)
{
#region Create Output layer
for (int i = 0; i < 5; i++) // Construct the output Layer
{
Perceptron per = new Perceptron(k);
OutPut.Add(per);
}
#endregion
#region Get_Cntres
// Get Random Centers For Clusters
Random rd = new Random();
for (int i = 0; i < k; i++)
{
cluster temp = new cluster();
temp.center = Training_data[rd.Next(0, Training_data.Count)];
Cluster.Add((temp));
KeyValuePair<int, float[]> ls = new KeyValuePair<int, float[]>();
sample s = new sample();
s.copysample(temp.center);
old_centers.Add(s);
}
#endregion
// old_centers[0].Value[0] = Training_data[rd.Next(0,Training_data.Count)].Value[0];
old_centers[0].features[0] = (float)-0.0002;
while (chek())
{
List<double> _out = new List<double>();
#region updateing_old_centers
for (int i = 0; i < k; i++)
{
Cluster[i].cluster_samples.Clear();
old_centers[i].copysample(Cluster[i].center);
}
#endregion
#region Clustring
//Clustering // Here's a trap
for (int i = 0; i < Training_data.Count; i++) // outer one
{
//for (int z = 0; z < Training_data[i].Count; z++)
{
_out.Clear();
for (int j = 0; j < k; j++)
{
_out.Add(distance(Training_data[i].features, Cluster[j].center.features));
}
Cluster[get_min(_out)].cluster_samples.Add(Training_data[i]);
}
}
#endregion
#region get new centers
for (int c = 0; c < k; c++)
{
double[] x = new double[128];
for (int i = 0; i < Cluster[c].cluster_samples.Count; i++)
{
for (int j = 0; j < 128; j++)
{
x[j] += Convert.ToDouble(Cluster[c].cluster_samples[i].features[j]);
}
}
if (Cluster[c].cluster_samples.Count != 0)
{
for (int sd = 0; sd < 128; sd++)
{
x[sd] /= Cluster[c].cluster_samples.Count;
}
// Cluster[c].center.;
for (int sd = 0; sd < 128; sd++)
{
Cluster[c].center.features[sd] = (float)(x[sd]);
}
}
}
#endregion
}
}
