CS 545 : Assignment 8

A Self-Organizing Map (SOM) is a type of artificial neural network used to transform a data set of vectors into a set of lower dimensional vectors. The applications of this are wider than one might expect, but this is the heart of its purpose. The data are usually some set of vectors {x ∈ Rn}, but a SOM can work with any set of vectors that have a well-defined distance metric. The canonical examples and demonstrations of SOMs reduce an n-dimensional data set to a 2-D data set. In section 2, I describe SOMs in more detail. In section 3, I present the traditional iterative algorithm used to construct Self-organizing maps. I include the R code that I use to generate the data, run the SOM algorithm iteratively, and display the results in section 4. Section 5 includes several examples of SOMs constructed from data sets in [0, 1] × [0, 1] × [0, 1] interpreted as color data (where each dimension is the intensity of one of red, blue, or green). Applications of SOMs are discussed in section 6, and conclusions are included in section 7