Distance matrix

A common strategy in clustering the units of the SOM is to calculate a distance matrix between the reference vectors and use a high value of the matrix as an indication of a cluster border [46, 26, 45]. In 3D visualization of such a matrix, the clusters will appear as ``valleys''. The problem then is how to determine which map units belong to a given cluster. For this agglomerative and devisive algorithms are typically used, e.g. in [32, 49]. Agglomerative algorithms usually have the following steps:

1. assign each map unit to its own cluster
2. calculate the distances between all clusters
3. join the two closest clusters
4. if only a user-defined number of clusters exists, end, otherwise iterate from step 2

In addition to distance some other joining criteria can be used. In contiguity-constrained clustering the clusters in step 3 are required to be adjacent [32].