ストリームデータのオンライン特徴抽出アルゴリズム―追加型主成分分析―

カテゴリ: 論文誌(論文単位)

グループ名: 【C】電子・情報・システム部門

発行日: 2012/01/01

タイトル(英語): Online Feature Extraction Algorithms for Data Streams―Incremental Principal Component Analysis―

著者名: 小澤　誠一（神戸大学大学院工学研究科）

著者名(英語): Seiichi Ozawa (Graduate School of Engineering, Kobe University)

キーワード: パターン認識，特徴抽出，追加学習，ストリームデータ　　pattern recognition，feature extraction，incremental learning，streaming data

要約(英語): Along with the development of the network technology and high-performance small devices such as surveillance cameras and smart phones, various kinds of multimodal information (texts, images, sound, etc.) are captured real-time and shared among systems through networks. Such information is given to a system as a stream of data. In a person identification system based on face recognition, for example, image frames of a face are captured by a video camera and given to the system for an identification purpose. Those face images are considered as a stream of data. Therefore, in order to identify a person more accurately under realistic environments, a high-performance feature extraction method for streaming data, which can be autonomously adapted to the change of data distributions, is solicited. In this review paper, we discuss a recent trend on online feature extraction for streaming data. There have been proposed a variety of feature extraction methods for streaming data recently. Due to the space limitation, we here focus on the incremental principal component analysis.

本誌: 電気学会論文誌C（電子・情報・システム部門誌） Vol.132 No.1 （2012） 特集：確率的最適化と機械学習の統計的設計と応用

本誌掲載ページ: 2023/06/13 p

原稿種別: 解説／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejeiss/132/1/132_1_6/_article/-char/ja/