Multichannel Singular Spectrum Analysis

Multichannel Singular Spectrum Analysis, or mSSA, is a powerful non-parametric technique for extracting temporally correlated signals from a set of input time series. The theory behind the work is described primarily in Weinberg & Petersen (2021). More recently, we published a study of an isolated disc-halo simulation using mSSA in Johnson, Petersen, Johnston, Weinberg (2023).

Project Overview

mSSA is a rich method that falls under the umbrella of dynamical data discovery. I'm interested in continuing to develop the methodology for use across simulations, as well as data. For now, we've used mSSA to identify several instances of nonlinear behaviour in simulations: bar slowdown, harmonic coupling between dipoles and quadrupoles, and velocity signatures.

Pedagogical Introduction

I've built a small pedagogical demonstration in pure Python, which has a RISE slideshow format. Click here. Please contact me if you are interested in the executable version of the notebook!