Functional Poisson approximation in Kantorovich-Rubinstein distance with applications to \(U\)-statistics and stochastic geometry (Q726803)

The authors study point processes which are constructed in the following way: a mass of 1 is assigned to every value a function takes for a \(k\)-tuple of points from the point process. This is related to \(U\)-statistics, which is the sum of the values for all possible tuples. It is shown that the Kantorovich-Rubenstein-distance of two Poisson processes is linked to the total variation distance of their intensity measures. This theorem is then applied to prove rates of convergence in several examples: the approximation of the distribution of a \(U\)-statistic by a Poisson distribution or an alpha stable distribution. As examples from stochastic geometry, the authors treat the approximation of random geometric graphs and of the midpoints of Poisson flats by a Poisson processes.