Recurrent switching linear dynamical systems

TL;DR

This paper introduces recurrent switching linear dynamical systems, a novel model that uncovers dynamic units and their switching conditions in complex time series data, enhancing interpretability over traditional models.

Contribution

The paper proposes a new class of models that jointly discover dynamic units and their switching conditions, improving interpretability and inference efficiency over existing SLDS models.

Findings

Model effectively uncovers dynamic units in complex data.

Switching behavior depends on observations or latent states.

Inference is made scalable with recent algorithmic advances.

Abstract

Many natural systems, such as neurons firing in the brain or basketball teams traversing a court, give rise to time series data with complex, nonlinear dynamics. We can gain insight into these systems by decomposing the data into segments that are each explained by simpler dynamic units. Building on switching linear dynamical systems (SLDS), we present a new model class that not only discovers these dynamical units, but also explains how their switching behavior depends on observations or continuous latent states. These "recurrent" switching linear dynamical systems provide further insight by discovering the conditions under which each unit is deployed, something that traditional SLDS models fail to do. We leverage recent algorithmic advances in approximate inference to make Bayesian inference in these models easy, fast, and scalable.