Introduction to SOM Toolbox

SOM Toolbox is a software library for Matlab 5 (version 5.2 at least) implementing the Self-Organizing Map (SOM) algorithm.

SOM_PAK is a public domain software package, written in C language for UNIX and PC environments. It is a semi-official implementation of the SOM algorithm.

Matlab has been steadily gaining popularity as the "language of scientific computing". For quite a while, the Matlab Neural Networks Toolbox has included a couple of functions that are related to the SOM. These are, however, primarily for demonstrations of the self-organization process and, as such, not sufficient for practical data analysis applications. As far as we know the SOM Toolbox is the first such software for Matlab.

So there seemed to be a need also for a Matlab implementation of the algorithms found in SOM_PAK. Moreover, we felt that Matlab might be well-suited for fast prototyping and customizing as Matlab employs a high-level programming language with strong support for graphics and visualization. It is because of these two reasons that we considered it worthwhile to develop the SOM Toolbox.

The SOM_PAK files can also be accessed with the Toolbox, so it is possible to first train the map with the SOM_PAK and then use the Toolbox for map visualization. Read more about SOM Toolbox vs. SOM_PAK >>>.

Highlights of the SOM Toolbox include the following:

• Modular programming style: the Toolbox code utilizes Matlab structures and the functions are constructed in a modular manner, which makes it convenient to tailor the code for each users' specific needs
• Component names, masks and normalizations: to facilate data mining process, the input vector components may be given names, and different kinds of (reversible) preprocessing operations can be defined for them. Also, the components may be masked, or weighted, according to their relative importance
• Batch or sequential training: in data analysis applications, the speed of training may be considerably improved by using the batch version. There are also other training variants, like supervised SOM.
• Map dimension: maps may be N-dimensional --- although visualization is not supported when N > 2
• Advanced graphics: building on the Matlab's strong graphics capabilities, attractive figures can be easily produced
• GUIs: there are also some graphical user interfaces, although the use of command line versions of the functions is strongly recommended