Ward clustering - Kelly method

Hello All,





I am using Ward algorithm implemented in JKlustor (3.1.3) to cluster a set of 8964 molecules. Before that, I run Kelly optimization program to have an idea on a number of possible clusters. How do you explain the following result :





Kelley Indexes for All Cluster Levels





level index


1 8964,0000


2 4059,8250


.


.


8961 8962,0000


8962 8963,0000


8963 8964,0000





Optimal number of clusters: -1 ??





Thank you for your support.





Best regards,





IsI

Hi!


This definitely a bug.


Could you send some example so I can reproduce it.


Thank you


Andras

Hi,





Attached the corresponding file. Thank you for your help.





Kind regards,


IsI

Hi!





I have found the bug and fixed it.


Thank you for the attachment.





Andras

Hi,





Thank you for fixing this bug. Where can I find the new version of the algorithm ?





Best regards,





IsI

Hi,





I have prepared a test jchem.jar for you which can be downloaded from here:


http://www.chemaxon.com/shared/jchem_test_2005_12_07/jchem.jar





Please note that this .jar was built from untested code under development, so I only recommend it for testing purposes.





Best regards,





Szilard