Adaptive sample set construction is a heuristic clustering method commonly used when the number of classses is unknown. This algorithm is simple and efficient: minimum computations are required and pattern samples are processed sequentially without the necessity of being stored. On the other hand, there are some drawbacks to the use of this algorithm: clustering results are dependent on the order of presentation and strong assumptions are required (such as that clusters should be tight and widely separated from one another).
The first cluster center is chosen arbitrarily. A pattern is assigned to a given cluster if the distance from the sample to the cluster is below tetha*tau. If that distance is greater than tau, a new cluster is created. No decision will be made if the pattern falls in the intermediate region.
After the training is complete, all patterns are assigned to the nearest class according to the minimum distance rule. Some patterns could be considered unclassified if their distances to all cluster centers are greater than tau.
Parameters