ROBUSTIZATION OF A LEARNING-METHOD FOR RBF NETWORKS

被引:32
作者
SANCHEZ, VD
机构
[1] German Aerospace Research Establishment, DLR Oberpfaffenhofen, Institute for Robotics and System Dynamics
关键词
FUNCTION APPROXIMATION; LEARNING FROM EXAMPLES; NETWORK SYNTHESIS; RADIAL BASIS FUNCTIONS; REGRESSION; ROBUST LEARNING METHODS;
D O I
10.1016/0925-2312(95)00000-V
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Outliers in a given training data set can cause substantial deterioration of the approximation realized by a neural network architecture. Robustization refers to the process of enhancing a learning method to deal with data containing outliers. The robustization of a learning method for training RBF networks for function approximation is presented leading to a novel efficient robust learning method for this architecture class. Its experimental evaluation confirms the expectations created from its theoretical foundations.
引用
收藏
页码:85 / 94
页数:10
相关论文
共 12 条
[1]  
Broomhead D. S., 1988, Complex Systems, V2, P321
[2]   A ROBUST BACK-PROPAGATION LEARNING ALGORITHM FOR FUNCTION APPROXIMATION [J].
CHEN, DS ;
JAIN, RC .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (03) :467-479
[3]   THE CHANGE-OF-VARIANCE CURVE AND OPTIMAL REDESCENDING M-ESTIMATORS [J].
HAMPEL, FR ;
ROUSSEEUW, PJ ;
RONCHETTI, E .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1981, 76 (375) :643-648
[4]  
HAMPEL FR, 1986, ROBUST STATISTICS
[5]  
Huber PJ, 1981, ROBUST STATISTICS
[6]  
Jokinen P. A., 1991, Neurocomputing, V3, P157, DOI 10.1016/0925-2312(91)90039-E
[7]   Fast Learning in Networks of Locally-Tuned Processing Units [J].
Moody, John ;
Darken, Christian J. .
NEURAL COMPUTATION, 1989, 1 (02) :281-294
[8]   NETWORKS FOR APPROXIMATION AND LEARNING [J].
POGGIO, T ;
GIROSI, F .
PROCEEDINGS OF THE IEEE, 1990, 78 (09) :1481-1497
[9]  
Press W. H., 1988, NUMERICAL RECIPES C
[10]   2ND DERIVATIVE DEPENDENT PLACEMENT OF RBF CENTERS [J].
SANCHEZ, VD .
NEUROCOMPUTING, 1995, 7 (03) :311-317