Average Linkage is computed wrongly

[GoogleCodeExporter]

cSize[c1] += cSize[c2];
...
else if (linkage == figue.AVERAGE_LINKAGE) {
var avg = ( cSize[c1] * distMatrix[c1][j] + cSize[c2] *        
distMatrix[c2][j])  / (cSize[c1] + cSize[j]) 
distMatrix[j][c1] = distMatrix[c1][j] = avg;

should be:

var avg = ( cSize[c1] * cSize[j] * distMatrix[c1][j] + cSize[c2] * cSize[j] * 
distMatrix[c2][j])  / ( (cSize[c1] + cSize[c2]) * cSize[j] ) 
distMatrix[j][c1] = distMatrix[c1][j] = avg;
....
cSize[c1] += cSize[c2]; // update cluster size later

Original issue reported on code.google.com by netge...@gmail.com on 16 Oct 2012 at 11:12

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Or eliminating the cSize[j] term we can write:

var avg = (cSize[c1] * distMatrix[c1][j] + cSize[c2] * distMatrix[c2][j])  / 
(cSize[c1] + cSize[c2/*j*/])  
...
cSize[c1] += cSize[c2]; // update cluster size after average linkage 
calculations

Original comment by netge...@gmail.com on 17 Oct 2012 at 10:59