Connectivity Graphs and Clustering with Similarity Functions

To construct a similarity graph we transform a given set x1,...,x n of data points with pairwise similarities s ij or distances d ij into a graph. There are several popular methods to construct similarity graphs [154]. The goal of constructing similarity graphs is to model the local neighborhood relationships between data points. In this section we review two popular methods to construct a similarity graph, and then we introduce a new algorithm that solves some problems that can not be solved by the others.