Prior to 2007

Self-Organizing Map for Web Search Results

Tuesday, April 11, 2000
AK218 - 3:00 p.m.
Sergey Makarov, Ph.D., ECE Department, WPI

Abstract

The Self-Organizing Map (SOM), or Kohonen's Map, is one of the most widely used neural network (NN) algorithms, with thousands of applications covered in the literature. Currently this method has been included in a large number of commercial and public doMain software packages. In this lecture, we will explain how this particular kind of neural network operates.

A SOM network provides an ordered 2D map of the collection: similar Documents lie near each other on the map. They form clusters, which may correspond to specific Research directions, companies, grants, etc. The relative position and magnitude of these clusters can be analyzed to determine the current status of the Research area. The SOM map is a nonlinear kind of mapping. In this sense, it enables to highlight some hidden and unexpected regularities. If the SOM map is periodically updated in time, we can see how different clusters grow up or dissipate. The temporal cluster dynamics may be used for prediction purposes.

April 11, 2000