論文

査読有り
2018年3月1日

Reconstructing latent dynamical noise for better forecasting observables

Chaos
  • Yoshito Hirata

28
3
記述言語
英語
掲載種別
研究論文(学術雑誌)
DOI
10.1063/1.4996043
出版者・発行元
American Institute of Physics Inc.

I propose a method for reconstructing multi-dimensional dynamical noise inspired by the embedding theorem of Muldoon et al. [Dyn. Stab. Syst. 13, 175 (1998)] by regarding multiple predictions as different observables. Then, applying the embedding theorem by Stark et al. [J. Nonlinear Sci. 13, 519 (2003)] for a forced system, I produce time series forecast by supplying the reconstructed past dynamical noise as auxiliary information. I demonstrate the proposed method on toy models driven by auto-regressive models or independent Gaussian noise.

リンク情報
DOI
https://doi.org/10.1063/1.4996043
URL
http://orcid.org/0000-0002-9245-2543

エクスポート
BibTeX RIS