Clustering Toolbox and Classification Software
Participants : Marc Csernel, Sergiu Chelcea, Francesco de Carvalho, Aicha El Golli, Brieuc Conan-Guez, Mihai Jurca, Yves Lechevallier [ co-correspondant ] , Brigitte Trousse [ co-correspondant ] .
For clustering, we maintained a clustering toolbox, written in C++ and Java, which groups clustering methods developed by the team over time, and uses the SOM library developed by M. Csernel. This library proposes a common data interface to every algorithm. This toolbox supports developers in integrating various classification methods and testing and comparing with other methods. Now it integrates various methods:
from AxIS Rocquencourt: 1) a partitionning clustering method on complex data tables called SCluster  , 2) an adapted version of the SOM on the dissimilarity tables called DSOM  (cf. section 6.3.1 ) and Div (in C++)  ;
two partitionning clustering methods on the dissimilarity tables: 1) CDis (in C++)  issued from a collaboration between AxIS Rocquencourt team and Recife University, Brazil and 2) CCClust (in C++) issued from a collaboration between AxIS Rocquencourt team and Recife University, Brazil;
2-3 AHC (in Java)  (cf. section 6.3.4 ) from AxIS Sophia Antipolis.
We developed a Web interface for this clustering toolbox for the following methods: SCluster, Div, Cdis, CCClust. 2-3 AHC is available as a Java applet which runs the hierarchies visualisation toolbox . The aim of this online interface is in a short term to allow other team members (and in the near future Internet users) to use these classification methods to process their own data via the Web. The Web interface is developed in C++, run on our Apache internal Web server.
For classification of functional data, we developed a functional Multi-Layer Perceptron Method called FNET (cf. section 6.3.2 ).