A Basis Function Consolidation Method of Radial Basis Function Networks

カテゴリ: 部門大会

論文No: SS1-3

グループ名: 【C】平成28年電気学会電子・情報・システム部門大会講演論文集

発行日: 2016/08/31

タイトル(英語): A Basis Function Consolidation Method of Radial Basis Function Networks

著者名: 清水 俊樹(千葉大学),岡本 卓(千葉大学),小圷 成一(千葉大学)

著者名(英語): Toshiki Shimizu|Takashi Okamoto|Seiichi Koakutsu

キーワード: ラジアル基底関数ネットワーク|機械学習|Radial basis function network|machine learning

要約(日本語): Radial basis function (RBF) networks is used for function approximation, classification and so on. Generally ,RBF networks having many basis functions has superb performance compared to not having many basis functions.However, RBF networks having many basis functions has problem that it take high calculation costs on learning and operation. For example, it will become a problem that offer service that construct RBF networks for each user and process large quantity of data by using those. Therefore, it is important that changing networks scale more appropriate with keeping performance. In this paper, we propose a method in order to consolidate basis functions without significant degradation in performance. Results of computational experiments indicate the validity of the proposed method.

PDFファイルサイズ: 268 Kバイト