Clustering Algorithm Based On Valley Seeking

The MATLAB function vseek implements a valley seeking clustering algorithm by Koontz and Fukunaga.

As suggested by the title, the valley seeking algorithm attempts to locate inter-cluster
boundaries in valleys, i.e., regions of minimal object density. Starting with an initial cluster
assignment, it iteratively reassigns objects according to a voting rule. The algorithm
terminates when an iteration produces no reassignments.
In each iteration each object is assigned to a new cluster based on the votes of its
neighbors, i.e., objects within a Euclidean distance less than or equal to a specied radius.
Each neighbor votes for the cluster to which it is currently assigned. Thus if a majority of
its neighbors are currently assigned to cluster m, then the object is assigned to cluster m
for the next round.