Pattern Recognition, 4th Edition

Chapter 15. Clustering Algorithms IV

15.1. Introduction

The clustering algorithms presented in the previous two chapters evolved along two distinct major philosophies. The current chapter presents categories of algorithms that cannot be included in either of the previous two families, and they stem from various ideas. The first one includes clustering algorithms based on graph theory concepts, such as the minimum spanning tree, the directed tree and spectral clustering. The second category includes competitive learning algorithms. The third category includes branch and bound algorithms. These schemes guarantee to provide globally optimal clustering, in terms of a prespecified optimality criterion, at the cost of increased computational requirements. ...