当システム
  詳細検索
Japanese | English 

Nara Women's University Digital Information Repository > 020 理学部 > 学外刊行物(理学部) > 雑誌 >

このアイテムのファイル:

ファイル 記述 サイズ形式
AA00704814V85N6_064001.pdf著者版889.67 kBAdobe PDF見る/開く
タイトル: Supervised Learning of Two-Layer Perceptron under the Existence of External Noise — Learning Curve of Boolean Functions of Two Variables in Tree-Like Architecture —
著者: Uezu, Tatsuya link image; Kiyokawa, Shuji
著者(別表記) : 上江洌, 達也; 清川, 修二
著者読み: うえず, たつや; きよかわ, しゅうじ
発行日: 2016年 6月15日
出版者: 日本物理学会:The Physical Society of Japan
引用: Tatsuya Uezu, and Shuji Kiyokawa: Journal of the Physical Society of Japan, Vol. 85, Issue. 6, 064001
抄録: We investigate the supervised batch learning of Boolean functions expressed by a two-layer perceptron with a tree-like structure. We adopt continuous weights (spherical model) and the Gibbs algorithm. We study the Parity and And machines and two types of noise, input and output noise, together with the noiseless case. We assume that only the teacher suffers from noise. By using the replica method, we derive the saddle point equations for order parameters under the replica symmetric (RS) ansatz. We study the critical value αC of the loading rate α above which the learning phase exists for cases with and without noise. We find that αC is nonzero for the Parity machine, while it is zero for the And machine. We derive the exponents β¯β¯ of order parameters expressed as (α−αC)β¯(α−αC)β¯ when α is near to αC. Furthermore, in the Parity machine, when noise exists, we find a spin glass solution, in which the overlap between the teacher and student vectors is zero but that between student vectors is nonzero. We perform Markov chain Monte Carlo simulations by simulated annealing and also by exchange Monte Carlo simulations in both machines. In the Parity machine, we study the de Almeida–Thouless stability, and by comparing theoretical and numerical results, we find that there exist parameter regions where the RS solution is unstable, and that the spin glass solution is metastable or unstable. We also study asymptotic learning behavior for large α and derive the exponents β^β^ of order parameters expressed as α−β^α−β^ when α is large in both machines. By simulated annealing simulations, we confirm these results and conclude that learning takes place for the input noise case with any noise amplitude and for the output noise case when the probability that the teacher’s output is reversed is less than one-half.
記述: 電子版公開より12ヶ月後の2017年6月15日公開予定。著作権は一般社団法人日本物理 学会(The Physical Society of Japan)が保有しています。
???metadata.dc.relation.doi???: http://dx.doi.org/10.7566/JPSJ.85.064001
URI: http://hdl.handle.net/10935/4310
ISSN: 00319015
出現コレクション:雑誌

このアイテムの引用には次の識別子を使用してください: http://hdl.handle.net/10935/4310

当システムに保管されているアイテムはすべて著作権により保護されています。

 

Valid XHTML 1.0! DSpace Software Copyright © 2002-2006 MIT and Hewlett-Packard - ご意見をお寄せください