Abstract

The Self-Organizing Map (SOM) is a popular and well-studied unsupervised learning technique.  Much work has been done recently on visualizing the results of the SOM algorithm, using static non-interactive approaches.  This paper presents two new SOM visualization methods, based on the grand tour and linked brushing.  These new methods use animation to show progress of the algorithm in the input space, and interactivity to let the user explore the link between the map in input space and feature space.  The new methods were implemented in Java using JSOMap , a new Java-based SOM package, and Orca, a Java-based data visualization package.  Several different data sets are used to demonstrate the new methods.

Documentation

Downloads

Instructions

To start the SOM visualization software, follow these steps:

1. download somviz.zip
2. unzip it - it will create a somviz directory containing all of the files
3. execute the following on a command line in the new somviz directory

java -cp .;jsomap.jar;orca.jar;jama.jar examples.SOMExample data/DATAFILE.txt

where DATAFILE.txt is one of the data files in the data directory.

Please refer to the user manual for information about how to use the SOM visualization.

Send mail to zcox@iastate.edu with questions or comments about this web site.
Last modified: February 28, 2002