par SOM: A Parallel Implementation of the Self-Organizing Map
                      Exploiting Cache Effects: 
      Making the SOM Fit for Interactive High-Performance Data Analysis 

             Andreas Rauber, Philipp Tomsich and Dieter Merkl 
    Institute of Software Technology, Vienna University of Technology 
              Favoritenstraße 9-11/188, A-1040 Wien, Austria 
                   {andi,phil,dieter}@ifs.tuwien.ac.at


Abstract:

A large number of applications has shown, that the self-organizing map 
is a prominent unsupervised neural network model for high-dimensional 
data analysis. However, the high execution times required to train 
the map put a limit to its use in many application domains, where 
either very large datasets are encountered and/or interactive 
response times are required. 

In order to provide interactive response times during data analysis 
we developed the par SOM, a software-based parallel implementation of 
the self-organizing map. Parallel execution reduces the training time 
to a large degree, with an even higher speedup obtained by using the 
resulting cache effects. We demonstrate the scalability of the parSOM 
system and the speed-up obtained on different architectures using an 
example from high-dimensional text data classification. 


-------------

A Parallel Implementation of the Self-Organizing Map Exploiting Cache Effects: 
Making the SOM Fit for Interactive High-Performance Data Analysis Andreas Rauber, Philipp Tomsich and Dieter Merkl 
In: Proceedings of the International Joint Conference on Neural Networks 2000 (IJCNN'2000), 24. - 27. 7. 2000, Como, Italy.