SELF: A framework for generation of patterns

摘要

This study presents a framework for generating patterns from data. Underlying the framework developed in this paper is the view that certain aspects of clustering should be based on a consideration of the physical pattern in its own right as distinguished from the pattern which arises from purely mathematical considerations. The viewpoint is further developed and illustrated. A principal conclusion of the study is that the discriminatory power of the non-dimensional functions can serve as a clustering technique of great generality. The implications of the proposed framework are discussed in relation to other existing methodologies of pattern-recognition.