Abstract

In this paper, we design algorithms for clustering locally asymptoticallyself-similar stochastic processes. We show a sufficient condition on thedissimilarity measure that leads to the consistency of the algorithms forclustering offline and online data settings, respectively. As an example ofapplication, clustering synthetic data sampled from multifractional Brownianmotions is provided.

 

Quick Read (beta)