Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:A1L-A
Session:
Number:A1L-A-6
Optimal Model Selection for Estimating Stochastic Koopman Modes
Wataru Kurebayashi, Sho Shirasaka, Hiroya Nakao,
pp.12-12
Publication Date:2017/12/4
Online ISSN:2188-5079
Summary:
Dynamic mode decomposition (DMD) is a method of modal extraction from time series data, which models the latent nonlinear dynamics underlying the data in terms of the dynamical systems theory. Recently, the extended DMD (EDMD) [M. Williams et al., 2015], which significantly widens the applicability of DMD, has been proposed. However, selecting an optimal setting of EDMD for given data is still an open question. In this study, we propose an algorithm to select the optimal hyperparameter of EDMD. The validity of our method is demonstrated by numerical experiments.