Key points are not available for this paper at this time.
The problem of organizing and partitioning large amounts of data into classes such that all data in one class will have similar properties is well known in pattern recognition research. The first step in the process, a cluster finding technique, involves grouping a large amount of data into clusters which must be detected and encoded so that automatic pattern recognition can take place. This paper describes a method for detecting and coding clusters. The principal advantages of this technique are that clusters need not be known a priori and no matrix inversion is required.
Mattson et al. (Thu,) studied this question.