Ensemble and Deep Learning Applications for Showcase Fault Analysis

カテゴリ: 全国大会

論文No: 3-100

グループ名: 【全国大会】平成29年電気学会全国大会論文集

発行日: 2017/03/05

タイトル(英語): Ensemble and Deep Learning Applications for Showcase Fault Analysis

著者名: Adamo Santana(Federal University of Para),福山 良和(明治大学),村上 賢哉(富士電機),松井 哲郎(富士電機)

著者名(英語): Adamo Santana(Federal University of Para),Yoshikazu Fukuyama(Meiji University),Kenya Murakami(Fuji Electric Co., LTD.),Tetsuro Matsui(Fuji Electric Co., LTD.)

キーワード: Deep Learning|Auto Encoders|Ensemble|Refrigeration Showcase|Fault Analysis|Machine Learning

要約(日本語): This paper considered the ensemble and deep AEs (which are called Competitive AE (CAEs) and Stacked AE (SAEs) respectively, for the problem of fault identification in refrigeration showcases. Differently from correlated approaches, the simulations were carried out with no prior data transformation, or dataset improvement, which emphasizes the robustness of the algorithms.While the evaluated models provided similar results, and the inherent trade-off would direct towards using the less complex approach (i.e. single AE) for the problem, the cumulative output style provided by CAEs might bring additional information for experts, more expressively highlighting out-of-control samples.

原稿種別: 英語

PDFファイルサイズ: 191 Kバイト