Float HierarchicalClustering::computeClusterVariance( const ClusterInfo &cluster, const MatrixFloat &data ){
VectorFloat mean(N,0);
VectorFloat std(N,0);
//Compute the mean
UINT numSamples = cluster.getNumSamplesInCluster();
for(UINT j=0; j<N; j++){
for(UINT i=0; i<numSamples; i++){
UINT index = cluster[i];
mean[j] += data[ index ][j];
}
mean[j] /= Float( numSamples );
}
//Compute the std dev
for(UINT j=0; j<N; j++){
for(UINT i=0; i<numSamples; i++){
std[j] += grt_sqr( data[ cluster[i] ][j] - mean[j] );
}
std[j] = grt_sqrt( std[j] / Float( numSamples-1 ) );
}
Float variance = 0;
for(UINT j=0; j<N; j++){
variance += std[j];
}
return variance/N;
}
Float HierarchicalClustering::computeClusterDistance( const ClusterInfo &clusterA, const ClusterInfo &clusterB ){
Float minDist = grt_numeric_limits< Float >::max();
const UINT numSamplesA = clusterA.getNumSamplesInCluster();
const UINT numSamplesB = clusterB.getNumSamplesInCluster();
//Find the minimum distance between the two clusters
for(UINT i=0; i<numSamplesA; i++){
for(UINT j=0; j<numSamplesB; j++){
if( distanceMatrix[ clusterA[i] ][ clusterB[j] ] < minDist ){
minDist = distanceMatrix[ clusterA[i] ][ clusterB[j] ];
}
}
}
return minDist;
}