Table P9.15 presents a summary of the renormalized SOM algorithm; a brief description of the algorithm is given in Section 9.3.

Compare the conventional and renormalized SOM algorithms, keeping in mind the following two issues:

1. the coding complexity involved in algorithmic implementation;

2. the computer time taken to do the training.

Illustrate the comparison between these two algorithms using data drawn from a uniform distribution inside a square and the following two network configurations:

(a) a one-dimensional lattice of 257 neurons;

(b) a one-dimensional lattice of 2,049 neurons

In both cases, start with an initial number of code vectors equal to two.