Self-Organizing Maps Applet

Tom Germano


This is a demonstration of how Self-Organizing Maps work. The high dimensional data that is being organized here is colors which are 3 dimensional since they are composed of 3 components: red, blue, and green. Assuming SOMs work, the final picture produced should have similar colors close to each and when the “Similarity SOM” option is toggled, there should be black lines separating colors which are not similar to each other.

There are three different ways to view the organization process. The second button from the left is used to run through the algorithm once and then stop, the button to the right of that will continue running through the algorithm, and the button all the way to the right will continue organizing the map without displaying at each iteration in order to quickly organize the map. Hitting the button to the left will stop the organizing at any time.

The button labeled “Reset” is used to initialize the weight values. To the left of that is a listbox containing three different methods of initializing the weights.  Below this is the option of viewing the actual colored SOM and the SOM displaying the similarities and differences in the map as described determining the quality of the map. In addition to this is a scrollbar used for increasing the number of iterations the program will execute before completion and a meter indicating the percentage of iterations completed. The more iterations used, the better the map.