/// <summary>
/// Initialize empty clusters with centroid values based on their distances from each other.
/// </summary>
/// <param name="clusterCount">Number of clusters to initialize.</param>
protected void InitEmptyKmeansClusters(uint clusterCount)
{
_clusters = new Dictionary <uint, Cluster>();
CentroidPoint firstcentroid = new CentroidPoint(PointDim);
firstcentroid.FillWithRandomValues(); //first center is random
_clusters.Add(firstcentroid.ID, new Cluster(firstcentroid));
for (uint i = 1; i < clusterCount; i++)
{
CentroidPoint centroid = new CentroidPoint(PointDim);
Point farthest = new Point(PointDim);
float maxDist = 0.0f;
foreach (ImagePoint point in _points)
{
float dist = 0.0f;
foreach (var cluster in _clusters)
{
dist += Point.SquaredEuclDist(cluster.Value.Centroid, point);
}
if (dist > maxDist)
{
maxDist = dist;
farthest = point;
}
}
centroid.CopyValuesFrom(farthest);
_clusters.Add(centroid.ID, new Cluster(centroid));
}
}
/// <summary>
/// Initialize empty clusters with centroid values randomly generated from their domain.
/// </summary>
/// <param name="clusterCount">Number of clusters to initialize.</param>
protected void InitEmptyClusters(uint clusterCount)
{
_clusters = new Dictionary <uint, Cluster>();
for (uint i = 0; i < clusterCount; i++)
{
CentroidPoint centroid = new CentroidPoint(PointDim);
centroid.FillWithRandomValues();
_clusters.Add(centroid.ID, new Cluster(centroid));
}
}
