Statistical Mechanics of On-line Node-perturbation Learning

• Item type: Trans(1)
• 公開日: 2011-01-26
• タイトル: Statistical Mechanics of On-line Node-perturbation Learning
• タイトル / 言語: en
• タイトル: Statistical Mechanics of On-line Node-perturbation Learning
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: オリジナル論文
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_6501
• 資源タイプ: journal article
• 著者所属: College of Industrial Technology, Nihon University
• 著者所属: Japan Science Technology Agency, ERATO Okanoya Emotional Information Project／Brain Science Institute, RIKEN／Graduate School of Frontier Science, The University of Tokyo
• 著者所属: Brain Science Institute, RIKEN／Japan Science Technology Agency, ERATO Okanoya Emotional Information Project
• 著者所属: Graduate School of Frontier Science, The University of Tokyo／Brain Science Institute, RIKEN／Japan Science Technology Agency, ERATO Okanoya Emotional Information Project
• 著者所属(英): en
• 著者所属(英): College of Industrial Technology, Nihon University
• 著者所属(英): en
• 著者所属(英): Japan Science Technology Agency, ERATO Okanoya Emotional Information Project / Brain Science Institute, RIKEN / Graduate School of Frontier Science, The University of Tokyo
• 著者所属(英): en
• 著者所属(英): Brain Science Institute, RIKEN / Japan Science Technology Agency, ERATO Okanoya Emotional Information Project
• 著者所属(英): en
• 著者所属(英): Graduate School of Frontier Science, The University of Tokyo / Brain Science Institute, RIKEN / Japan Science Technology Agency, ERATO Okanoya Emotional Information Project
• 著者名: Kazuyuki, Hara; Kentaro, Katahira; Kazuo, Okanoya; Masato, Okada
• 著者名(英): Kazuyuki, Hara; Kentaro, Katahira; Kazuo, Okanoya; Masato, Okada
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Node-perturbation learning (NP-learning) is a kind of statistical gradient descent algorithm that estimates the gradient of an objective function through application of a small perturbation to the outputs of the network. It can be applied to problems where the objective function is not explicitly formulated, including reinforcement learning. In this paper, we show that node-perturbation learning can be formulated as on-line learning in a linear perceptron with noise, and we can derive the differential equations of order parameters and the generalization error in the same way as for the analysis of learning in a linear perceptron through statistical mechanical methods. From analytical results, we show that cross-talk noise, which originates in the error of the other outputs, increases the generalization error as the output number increases.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Node-perturbation learning (NP-learning) is a kind of statistical gradient descent algorithm that estimates the gradient of an objective function through application of a small perturbation to the outputs of the network. It can be applied to problems where the objective function is not explicitly formulated, including reinforcement learning. In this paper, we show that node-perturbation learning can be formulated as on-line learning in a linear perceptron with noise, and we can derive the differential equations of order parameters and the generalization error in the same way as for the analysis of learning in a linear perceptron through statistical mechanical methods. From analytical results, we show that cross-talk noise, which originates in the error of the other outputs, increases the generalization error as the output number increases.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA11464803
• 書誌情報: 情報処理学会論文誌数理モデル化と応用（TOM）; 巻 4, 号 1, p. 72-81, 発行日 2011-01-26
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 1882-7780
• 出版者 / 言語: ja
• 出版者: 情報処理学会