Vietoris-Rips Persistent Homology (1905.06071v1)

Abstract: Persistence diagrams are useful displays that give a summary information regarding the topological features of some phenomenon. Usually, only one persistence diagram is available, and replicated persistence diagrams are needed for statistical inference. One option for generating these replications is to fit a distribution for the points on the persistence diagram. The type of the relevant distribution depends on the way the persistence diagram is builded. There are two approaches for building the persistence diagram, one is based on the Vietoris-Rips complex, and the second is based on some fitted function such as the kernel density estimator. The two approaches yield a two dimensional persistence diagram, where the coordinates of each point are the 'birth' and 'death' times. For the first approach, however, the 'birth' time is zero for all the points that present the connected components of the phenomenon. In this paper we examine the distribution of the connected components when the persistence diagram is based on Vietoris-Rips complex. In addition, we study the behaviour of the connected components when the phenomenon is measured with noise.