Pattern Recognition, 4th Edition

14.5. Hard Clustering Algorithms

In this section we return to the world where each vector belongs exclusively to a single cluster. This is why such schemes are called hard or crisp clustering algorithms. It turns out that some of the most well-known and widely used clustering algorithms fall into this category. Our starting point is the assumption that the membership coefficients u^ij are either 1 or 0. Moreover, they are 1 for one cluster, C^j, and zero for all the others, C^k, k ≠ j, that is,(14.79)and(14.80)This situation may be seen as a special ...