Applications of SOM magnification to data mining

Magnification in Self-Organizing Maps refers to the functional relationship between the density of the SOM weights in input space, and the density of the input space. The explicit magnification control scheme proposed by Bauer, Der and Herrmann (1) in 1996 opened the possibility to achieve specific magnifications that have attractive properties for data mining. However, the theoretical support only extends to 1- and 2- dimensional data with independent dimensions. This paper studies the scope of validity of the magnification control approach in hope to justify its application to real, high-dimensional data, which do not fall in the categories supported by the theory. We show encouraging results on synthetic as well as on real data.

[1]  Erzsébet Merényi,et al.  Forbidden Magnification? I , 2004, ESANN.

[2]  Helge Ritter.  Asymptotic level density for a class of vector quantization processes , 1991, IEEE Trans. Neural Networks.

[3]  K. Schulten,et al.  On the stationary state of Kohonen's self-organizing sensory mapping , 2004, Biological Cybernetics.

[4]  Erzsébet Merényi,et al.  Forbidden magnification? II , 2004, ESANN.

[5]  Duane DeSieno,et al.  Adding a conscience to competitive learning , 1988, IEEE 1988 International Conference on Neural Networks.

[6]  Ralf Der,et al.  Controlling the Magnification Factor of Self-Organizing Feature Maps , 1996, Neural Computation.

[7]  Teuvo Kohonen,et al.  The self-organizing map , 1990, Neurocomputing.

[8]  Erzsébet Merényi,et al.  Mapping surface materials on Mars from Mars Pathfinder spectral images with HYPEREYE , 2004, International Conference on Information Technology: Coding and Computing, 2004. Proceedings. ITCC 2004..