Numerical Cruncher

Clustering

Batchelor & Wilkins' Algorithm
(maximum distance algorithm)

Batchelor and Wilkins' Alogorithm is a clustering algorithm used when the number of classes is unknown, as the adaptive sample set construction method.

Parameters

• f: Fraction of the average interset distance (average distance between clusters)

Clustering algorithm

• First cluster: Randomly selected pattern.
  
• Second cluster: Pattern sample furthest from the first cluster.
  
• While new clusters are being created, get the furthest pattern from the actual set of clusters (maximum of the minimum distances from the patterns to the clusters). If the distance from the chosen pattern to the set of clusters is above a fraction f of the average interset distance, form a new cluster with the selected pattern.
  
• Assign each pattern to its nearest cluster.