[4P2-J-3] Data mining: spaciotemporal data

[4P2-J-3-02] Time-Varying Dynamic Mode Decomposition

〇Naoya Takeishi¹ (1. RIKEN)

Keywords:dynamic mode decomposition, time-series data, nonlinear dynamical systems

In this talk, I introduce a data-driven method for modal decomposition of possibly time-varying dynamical systems. The formulation of the method is based on the probabilistic version of dynamic mode decomposition (DMD). While DMD and its many variants extract modes with fixed frequencies and decay rates, the proposed method aims to extract modes that have time-varying frequencies and decay rates. I show, with the results of preliminary experiments, that the proposed method works properly on time-varying dynamics.