The Self-Organising Map is a popular unsupervised neural network model which has successfully been used for analysing various kinds of data. The SOM performs both a vector quantisation, i.e. finding of prototypical representatives of the data, such as in k-means clustering, as well as a vector projection, that performs topology-preserving mapping from a high-dimensional input space to a normally two-dimensional output space, the map.
The Java SOMToolbox is an open-source implementation in Java, that allows you to easily train self-organising maps, and analyse them with an advanced viewer application, which implements a large range of different visualisations and quality measures of the SOM. These allow in-depth analysis and evaluation of the trained maps and the characteristics of the data, resulting in a powerful tool for data mining.
The Java SOMToolbox is developed at the Institute of Software Technology and Interactive System at the Vienna University of Technology and licensed under the Apache License, Version 2.0.
Contact: Rudolf Mayer and Andreas Rauber.
