ABSTRACT

After reading through this chapter, the reader will know about the most popular similarity clustering approaches including

1. Sources of similarity data including networks and graphs, as well as kernel, or affinity, transformations

2. Summary clustering with background noise subtracted, including a. Uniform clustering in which a noise threshold, or the similarity

shift, value is subtracted b. Modularity clustering in which random interaction noise is sub-

tracted 3. Additive clustering 4. Spectral clustering including Laplacian, and pseudo-inverse Lapla-

cian, transformation of similarity data 5. Consensus clustering including

a. Consensus ensemble data recovery clustering b. Combined consensus data recovery clustering c. Concordant partition and a drawback of its criterion

6. Graph-theoretic clustering including connected components, Maximum Spanning Tree and Single Linkage clustering

Additive cluster A subset of the entity set with a positive intensityweight assigned to it. A set of additive clusters generates a similarity matrix so that every two entities are assigned with a similarity score which is equal to the sum of the intensity weights of those clusters that contain both of the entities.