Effective training methods for function localization neural networks

Takafumi Sasakawa; Jinglu Hu; Katsunori Isono; Kotaro Hirasawa

doi:10.1109/ijcnn.2006.247154

Effective training methods for function localization neural networks

Takafumi Sasakawa^*, Jinglu Hu, Katsunori Isono, Kotaro Hirasawa

^*Corresponding author for this work

Graduate School of Information, Production and Systems

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Citation (Scopus)

Abstract

Inspired by Hebb's cell assembly theory about how the brain worked, we have developed a function localization neural network (FLNN). The main part of a FLNN is structurally the same as an ordinary feedforward neural network, but it is considered to consist of several overlapping modules, which are switched according to input patterns. A FLNN constructed in this way has been shown to have better representation ability than an ordinary neural network. However, BP training algorithm for such FLNN is very easy to get stuck at a local minimum. In this paper, we mainly discuss the methods for improving BP training of the FLNN by utilizing the structural property of the network. Two methods are proposed. Numerical simulations are used to show the effectiveness of the improved BP training methods.

Original language	English
Title of host publication	International Joint Conference on Neural Networks 2006, IJCNN '06
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	4785-4790
Number of pages	6
ISBN (Print)	0780394909, 9780780394902
DOIs	https://doi.org/10.1109/ijcnn.2006.247154
Publication status	Published - 2006 Jan 1
Event	International Joint Conference on Neural Networks 2006, IJCNN '06 - Vancouver, BC, Canada Duration: 2006 Jul 16 → 2006 Jul 21

Publication series

Name	IEEE International Conference on Neural Networks - Conference Proceedings
ISSN (Print)	1098-7576

Conference

Conference	International Joint Conference on Neural Networks 2006, IJCNN '06
Country/Territory	Canada
City	Vancouver, BC
Period	06/7/16 → 06/7/21

ASJC Scopus subject areas

Software

Access to Document

10.1109/ijcnn.2006.247154

Cite this

Sasakawa, T., Hu, J., Isono, K., & Hirasawa, K. (2006). Effective training methods for function localization neural networks. In International Joint Conference on Neural Networks 2006, IJCNN '06 (pp. 4785-4790). Article 1716764 (IEEE International Conference on Neural Networks - Conference Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ijcnn.2006.247154

Effective training methods for function localization neural networks. / Sasakawa, Takafumi; Hu, Jinglu; Isono, Katsunori et al.
International Joint Conference on Neural Networks 2006, IJCNN '06. Institute of Electrical and Electronics Engineers Inc., 2006. p. 4785-4790 1716764 (IEEE International Conference on Neural Networks - Conference Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Sasakawa, T, Hu, J, Isono, K & Hirasawa, K 2006, Effective training methods for function localization neural networks. in International Joint Conference on Neural Networks 2006, IJCNN '06., 1716764, IEEE International Conference on Neural Networks - Conference Proceedings, Institute of Electrical and Electronics Engineers Inc., pp. 4785-4790, International Joint Conference on Neural Networks 2006, IJCNN '06, Vancouver, BC, Canada, 06/7/16. https://doi.org/10.1109/ijcnn.2006.247154

Sasakawa T, Hu J, Isono K, Hirasawa K. Effective training methods for function localization neural networks. In International Joint Conference on Neural Networks 2006, IJCNN '06. Institute of Electrical and Electronics Engineers Inc. 2006. p. 4785-4790. 1716764. (IEEE International Conference on Neural Networks - Conference Proceedings). doi: 10.1109/ijcnn.2006.247154

@inproceedings{0650463d1b224360af9780ca1bb38c7d,

title = "Effective training methods for function localization neural networks",

abstract = "Inspired by Hebb's cell assembly theory about how the brain worked, we have developed a function localization neural network (FLNN). The main part of a FLNN is structurally the same as an ordinary feedforward neural network, but it is considered to consist of several overlapping modules, which are switched according to input patterns. A FLNN constructed in this way has been shown to have better representation ability than an ordinary neural network. However, BP training algorithm for such FLNN is very easy to get stuck at a local minimum. In this paper, we mainly discuss the methods for improving BP training of the FLNN by utilizing the structural property of the network. Two methods are proposed. Numerical simulations are used to show the effectiveness of the improved BP training methods.",

author = "Takafumi Sasakawa and Jinglu Hu and Katsunori Isono and Kotaro Hirasawa",

year = "2006",

month = jan,

day = "1",

doi = "10.1109/ijcnn.2006.247154",

language = "English",

isbn = "0780394909",

series = "IEEE International Conference on Neural Networks - Conference Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "4785--4790",

booktitle = "International Joint Conference on Neural Networks 2006, IJCNN '06",

note = "International Joint Conference on Neural Networks 2006, IJCNN '06 ; Conference date: 16-07-2006 Through 21-07-2006",

}

TY - GEN

T1 - Effective training methods for function localization neural networks

AU - Sasakawa, Takafumi

AU - Hu, Jinglu

AU - Isono, Katsunori

AU - Hirasawa, Kotaro

PY - 2006/1/1

Y1 - 2006/1/1

N2 - Inspired by Hebb's cell assembly theory about how the brain worked, we have developed a function localization neural network (FLNN). The main part of a FLNN is structurally the same as an ordinary feedforward neural network, but it is considered to consist of several overlapping modules, which are switched according to input patterns. A FLNN constructed in this way has been shown to have better representation ability than an ordinary neural network. However, BP training algorithm for such FLNN is very easy to get stuck at a local minimum. In this paper, we mainly discuss the methods for improving BP training of the FLNN by utilizing the structural property of the network. Two methods are proposed. Numerical simulations are used to show the effectiveness of the improved BP training methods.

AB - Inspired by Hebb's cell assembly theory about how the brain worked, we have developed a function localization neural network (FLNN). The main part of a FLNN is structurally the same as an ordinary feedforward neural network, but it is considered to consist of several overlapping modules, which are switched according to input patterns. A FLNN constructed in this way has been shown to have better representation ability than an ordinary neural network. However, BP training algorithm for such FLNN is very easy to get stuck at a local minimum. In this paper, we mainly discuss the methods for improving BP training of the FLNN by utilizing the structural property of the network. Two methods are proposed. Numerical simulations are used to show the effectiveness of the improved BP training methods.

UR - http://www.scopus.com/inward/record.url?scp=40649128098&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=40649128098&partnerID=8YFLogxK

U2 - 10.1109/ijcnn.2006.247154

DO - 10.1109/ijcnn.2006.247154

M3 - Conference contribution

AN - SCOPUS:40649128098

SN - 0780394909

SN - 9780780394902

T3 - IEEE International Conference on Neural Networks - Conference Proceedings

SP - 4785

EP - 4790

BT - International Joint Conference on Neural Networks 2006, IJCNN '06

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - International Joint Conference on Neural Networks 2006, IJCNN '06

Y2 - 16 July 2006 through 21 July 2006

ER -

Effective training methods for function localization neural networks

Abstract

Publication series

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this