Resisting the influence of outliers in radial basis function neural networks

Jea Rong Tsai, Pau-Choo Chung, Chein I. Chang

Research output: Contribution to conferencePaper

1 Citation (Scopus)

Abstract

Radial basis function (RBF) neural networks were shown to be a promising network model in function approximation. Training an RBF network is usually approached based on a least square criterion, accompanied with an adaptive growing technique to determine the optimal size of network. With this approach, two problems usually arise when the training patterns contain outliers. Firstly, the least square would cause the network to incorrectly interpolate the outliers. Secondly, because of the interference of outliers, the number of nodes determined by traditional growing algorithm will stuck at a certain number, causing that the proper network size cannot be reached. In order to cope with the first problem, this paper proposes a method to construct a robust criterion function to replace with the least square criterion. For solving the second problem, the paper introduces a memory mechanism into the adaptive growing technique to restrain the influence of outliers. Simulation results indicate that the robust criterion function obtained using our method can effectively reduce the influence of outliers. Furthermore, with the incorporation of memory mechanism, a better size of network can be obtained.

Original languageEnglish
Pages42-51
Number of pages10
Publication statusPublished - 1996 Jan 1
EventProceedings of the 1996 IEEE Signal Processing Society Workshop - Kyota, Jpn
Duration: 1996 Sep 41996 Sep 6

Other

OtherProceedings of the 1996 IEEE Signal Processing Society Workshop
CityKyota, Jpn
Period96-09-0496-09-06

Fingerprint

Neural networks
Data storage equipment
Radial basis function networks

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Cite this

Tsai, J. R., Chung, P-C., & Chang, C. I. (1996). Resisting the influence of outliers in radial basis function neural networks. 42-51. Paper presented at Proceedings of the 1996 IEEE Signal Processing Society Workshop, Kyota, Jpn, .
Tsai, Jea Rong ; Chung, Pau-Choo ; Chang, Chein I. / Resisting the influence of outliers in radial basis function neural networks. Paper presented at Proceedings of the 1996 IEEE Signal Processing Society Workshop, Kyota, Jpn, .10 p.
@conference{2f5fcfad103c47ff8dba75abd0012abe,
title = "Resisting the influence of outliers in radial basis function neural networks",
abstract = "Radial basis function (RBF) neural networks were shown to be a promising network model in function approximation. Training an RBF network is usually approached based on a least square criterion, accompanied with an adaptive growing technique to determine the optimal size of network. With this approach, two problems usually arise when the training patterns contain outliers. Firstly, the least square would cause the network to incorrectly interpolate the outliers. Secondly, because of the interference of outliers, the number of nodes determined by traditional growing algorithm will stuck at a certain number, causing that the proper network size cannot be reached. In order to cope with the first problem, this paper proposes a method to construct a robust criterion function to replace with the least square criterion. For solving the second problem, the paper introduces a memory mechanism into the adaptive growing technique to restrain the influence of outliers. Simulation results indicate that the robust criterion function obtained using our method can effectively reduce the influence of outliers. Furthermore, with the incorporation of memory mechanism, a better size of network can be obtained.",
author = "Tsai, {Jea Rong} and Pau-Choo Chung and Chang, {Chein I.}",
year = "1996",
month = "1",
day = "1",
language = "English",
pages = "42--51",
note = "Proceedings of the 1996 IEEE Signal Processing Society Workshop ; Conference date: 04-09-1996 Through 06-09-1996",

}

Tsai, JR, Chung, P-C & Chang, CI 1996, 'Resisting the influence of outliers in radial basis function neural networks' Paper presented at Proceedings of the 1996 IEEE Signal Processing Society Workshop, Kyota, Jpn, 96-09-04 - 96-09-06, pp. 42-51.

Resisting the influence of outliers in radial basis function neural networks. / Tsai, Jea Rong; Chung, Pau-Choo; Chang, Chein I.

1996. 42-51 Paper presented at Proceedings of the 1996 IEEE Signal Processing Society Workshop, Kyota, Jpn, .

Research output: Contribution to conferencePaper

TY - CONF

T1 - Resisting the influence of outliers in radial basis function neural networks

AU - Tsai, Jea Rong

AU - Chung, Pau-Choo

AU - Chang, Chein I.

PY - 1996/1/1

Y1 - 1996/1/1

N2 - Radial basis function (RBF) neural networks were shown to be a promising network model in function approximation. Training an RBF network is usually approached based on a least square criterion, accompanied with an adaptive growing technique to determine the optimal size of network. With this approach, two problems usually arise when the training patterns contain outliers. Firstly, the least square would cause the network to incorrectly interpolate the outliers. Secondly, because of the interference of outliers, the number of nodes determined by traditional growing algorithm will stuck at a certain number, causing that the proper network size cannot be reached. In order to cope with the first problem, this paper proposes a method to construct a robust criterion function to replace with the least square criterion. For solving the second problem, the paper introduces a memory mechanism into the adaptive growing technique to restrain the influence of outliers. Simulation results indicate that the robust criterion function obtained using our method can effectively reduce the influence of outliers. Furthermore, with the incorporation of memory mechanism, a better size of network can be obtained.

AB - Radial basis function (RBF) neural networks were shown to be a promising network model in function approximation. Training an RBF network is usually approached based on a least square criterion, accompanied with an adaptive growing technique to determine the optimal size of network. With this approach, two problems usually arise when the training patterns contain outliers. Firstly, the least square would cause the network to incorrectly interpolate the outliers. Secondly, because of the interference of outliers, the number of nodes determined by traditional growing algorithm will stuck at a certain number, causing that the proper network size cannot be reached. In order to cope with the first problem, this paper proposes a method to construct a robust criterion function to replace with the least square criterion. For solving the second problem, the paper introduces a memory mechanism into the adaptive growing technique to restrain the influence of outliers. Simulation results indicate that the robust criterion function obtained using our method can effectively reduce the influence of outliers. Furthermore, with the incorporation of memory mechanism, a better size of network can be obtained.

UR - http://www.scopus.com/inward/record.url?scp=0029712059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0029712059&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:0029712059

SP - 42

EP - 51

ER -

Tsai JR, Chung P-C, Chang CI. Resisting the influence of outliers in radial basis function neural networks. 1996. Paper presented at Proceedings of the 1996 IEEE Signal Processing Society Workshop, Kyota, Jpn, .