Extraction of latent structure by sparse modeling

2017 Fiscal Year Final Research Report

• Project Area: Initiative for High-Dimensional Data-Driven Science through Deepening of Sparse Modeling
• Project/Area Number: 25120009
• Research Category: Grant-in-Aid for Scientific Research on Innovative Areas (Research in a proposed research area)
• Allocation Type: Single-year Grants
• Review Section: Complex systems
• Research Institution: The University of Tokyo
• Principal Investigator: Okada Masato 東京大学, 大学院新領域創成科学研究科, 教授 (90233345)
• Co-Investigator(Kenkyū-buntansha): 田中 和之 東北大学, 情報科学研究科, 教授 (80217017) ; 村田 昇 早稲田大学, 理工学術院, 教授 (60242038) ; 井上 真郷 早稲田大学, 理工学術院, 教授 (70376953) ; 永田 賢二 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 主任研究員 (10556062)
• Project Period (FY): 2013-06-28 – 2018-03-31
• Keywords: データ駆動科学 / 潜在構造抽出 / スペクトル分解 / ブラインドセンシング / スパースDMD / 全状態探索 / ES-DoS

Outline of Final Research Achievements

The sparse modeling team (B01-2) sets three tasks. Task 1 is applications of Bayesian spectral decomposition method to actual data. We developed a noise variance estimation method and a fast calculation method using L1 regularization and verified its effectiveness with actual data. In task 2, we developed a basis estimation and selection method using Sp-DMD for time series data, and applied it to actual data and verified its effectiveness. In task 3, a method of evaluating the appropriateness of the basis combination using an exhaustive search was developed, and this method was applied to actual data and the effectiveness was verified. Through research on these three tasks, we developed a universal method to extract latent structures using SpM and verified its effectiveness by actual data.

Free Research Field

高次元データ駆動科学