Study on optimization of hyperparameters in higher-order Granger reservoir computing for structural inference

カテゴリ：部門大会

論文No：SS1-2

グループ名：【C】2025年電気学会電子・情報・システム部門大会

発行日：2025/8/20

タイトル(英語)：Study on optimization of hyperparameters in higher-order Granger reservoir computing for structural inference

著者名：大槻 怜央（東京大学）,李 彬（東京大学）,杉野 正和（東京大学）,榛葉 健太（東京大学）,小谷 潔（東京大学）

著者名(英語)： Reo Otsuki (The University of Tokyo),Bin Li (The University of Tokyo),Masato Sugino (The University of Tokyo),Kenta Shimba (The University of Tokyo),Kiyoshi Kotani (The University of Tokyo)

キーワード：リザバー計算,力学系,因果推論,システム同定,reservoir computing,dynamical system,causal inference,system identification

要約(日本語)：Physiological and neural systems interact dynamically, giving rise to diverse states and functions. Their nonlinear, collective and emergent behaviors make it challenging to discover the underlying interactions. Recently, higher-order Granger reservoir computing (HoGRC) has been developed to simultaneously infer connectivity structure and predict the future time series. However, previous studies have not evaluated how the accuracy of structural depends on hyperparameters of HoGRC. To address this gap, we assess the effectiveness of HoGRC with a coupled chaotic system.

本誌掲載ページ：1767-1769p

原稿種別：英語

PDFファイルサイズ：387Kバイト