Regularized dynamic mode decomposition algorithm for time sequence predictions

Dynamic mode decomposition (DMD) aims at extracting intrinsic mechanisms in a time sequence via linear recurrence relation of its observables, thereby predicting later terms in the sequence. Stability is a major concern in DMD predictions. We adopt a regularized form and propose a Regularized DMD (R...

Full description

Saved in:
Bibliographic Details
Main Authors: Xiaoyang Xie, Shaoqiang Tang
Format: Article
Language:English
Published: Elsevier 2024-09-01
Series:Theoretical and Applied Mechanics Letters
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2095034924000667
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Dynamic mode decomposition (DMD) aims at extracting intrinsic mechanisms in a time sequence via linear recurrence relation of its observables, thereby predicting later terms in the sequence. Stability is a major concern in DMD predictions. We adopt a regularized form and propose a Regularized DMD (ReDMD) algorithm to determine the regularization parameter. This leverages stability and accuracy. Numerical tests for Burgers’ equation demonstrate that ReDMD effectively stabilizes the DMD prediction while maintaining accuracy. Comparisons are made with the truncated DMD algorithm.
ISSN:2095-0349