Training data selection method for generalization by multilayer neural networks

Training data selection method for generalization by multilayer neural networks

フォーマット:

論文

責任表示:

Hara, Kazuyuki ; Nakayama, Kenji

言語:

英語

出版情報:

1998-03-01

著者名:

掲載情報:

IEICE Trans. Fundamentals

ISSN:

0916-8508

巻:

E81-A

通号:

3

開始ページ:

374

終了ページ:

381

バージョン:

publisher

概要:

金沢大学大学院自然科学研究科知能情報・数理<br />A training data selection method is proposed for multilayer neural networks (MLNNs). This method selects a small number of the training data, which guarantee both generalization and fast training of the MLNN … s applied to pattern classification. The generalization will be satisfied using the data locate close to the boundary of the pattern classes. However, if these data are only used in the training, convergence is slow. This phenomenon is analyzed in this paper. Therefore, in the proposed method, the MLNN is first trained using some number of the data, which are randomly selected (Step 1). The data, for which the output error is relatively large, are selected. Furthermore, they are paired with the nearest data belong to the different class. The newly selected data are further paired with the nearest data. Finally, pairs of the data, which locate close to the boundary, can be found. Using these pairs of the data, the MLNNs are further trained (Step 2). Since, there are some variations to combine Steps 1 and 2, the proposed method can be applied to both off-line and on-line training. The proposed method can reduce the number of the training data, at the same time, can hasten the training. Usefulness is confirmed through computer simulation. 続きを見る

URL:

http://hdl.handle.net/2297/5654

タイトル・著者・出版者が同じ資料

詳細

受理日:	1998-03-01

類似資料:

1

2

3

4

5

6

7

8

9

10

11

12

1 論文 Selection of minimum training sata for generalization and on-line training by multilayer neural networks Hara, Kazuyuki, Nakayama, Kenji IEEE(Institute of Electrical and Electronics Engineers)	7 論文 Comparison of activation functions in multilayer neural network for pattern classification Hara, Kazuyuki, Nakayama, Kenji Institute of Electrical and Electronics Engineers (IEEE)
2 論文 A training data selection in on-line training for multilayer neural networks Hara, Kazuyuki, Nakayama, Kenji Institute of Electrical and Electronics Engineers (IEEE)	8 論文 Comparison of activation functions in multilayer neural network for pattern classification Hara, Kazuyuki, Nakayama, Kenji Springer-Verlag / ICANN'94
3 論文 Multi-frequency signal classification by multilayer neural networks and linear filter methods Hara, Kazuyuki, Nakayama, Kenji IEICE transactions on fundamentals of electronics, communications and computer	9 論文 Effects of activation functions in multilayer neural network for noisy pattern classification Hara, Kazuyuki, Nakayama, Kenji IEEE-INNS
4 論文 Signal classification based on frequency analysis using multilayer neural network with limited data and computations Hara, Kazuyuki, Nakayama, Kenji IEEE(Institute of Electrical and Electronics Engineers)	10 論文 Effects of activation functions in multilayer neural network for noisy pattern classification Hara, Kazuyuki, Nakayama, Kenji Institute of Electrical and Electronics Engineers (IEEE)
5 論文 Signal classification based on frequency analysis using multilayer neural network with limited data and computations Hara, Kazuyuki, Nakayama, Kenji IEEE = Institute of Electrical and Electronics Engineers	11 論文 Superresolution of multi-frequency signals using multilayer neural network supervised by backpropagation algorithm Hara, Kazuyuki, Nakayama, Kenji IEEE(Institute of Electrical and Electronics Engineers)
6 論文 Classification of multi-frequency signals with random noise using multilayer neural networks Hara, Kazuyuki, Nakayama, Kenji IEEE(Institute of Electrical and Electronics Engineers)	12 論文 High resolution of multi-frequencies using multilayer neural networks trained by back-propagation algorithm Hara, Kazuyuki, Nakayama, Kenji Institute of Electrical and Electronics Engineers (IEEE)