Nonparametric Bayesian identification of jump systems with sparse dependencies

Abstract

Many nonlinear dynamical phenomena can be effectively modeled by a system that switches among a set of conditionally linear dynamical modes. We consider two such Markov jump linear systems: the switching linear dynamical system (SLDS) and the switching vector autoregressive (S-VAR) process. In this paper, we present a nonparametric Bayesian approach to identifying an unknown number of persistent, smooth dynamical modes by utilizing a hierarchical Dirichlet process prior. We additionally employ automatic relevance determination to infer a sparse set of dynamic dependencies. The utility and flexibility of our models are demonstrated on synthetic data and a set of honey bee dances.