Statistical analysis of a two-layer backpropagation algorithm used for modeling nonlinear memoryless channels: The single neuron case

被引:17
作者
Bershad, NJ [1 ]
Ibnkahla, M [1 ]
Castanie, F [1 ]
机构
[1] INST NATL POLYTECH TOULOUSE,ENSEEIHT GAPSE,F-31077 TOULOUSE,FRANCE
关键词
D O I
10.1109/78.558493
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 [电气工程]; 0809 [电子科学与技术];
摘要
Neural networks have been used for modeling the nonlinear characteristics of memoryless nonlinear channels using backpropagation (BP) learning with experimental training data. In order to better understand this neural network application, this paper studies the transient and convergence properties of a simplified two-layer neural network that uses the BP algorithm and is trained with zero mean Gaussian data. The paper studies the effects of the neural net structure, weights, initial conditions, and algorithm step size on the mean square error (MSE) of the neural net approximation. The performance analysis is based on the derivation of recursions for the mean weight update that can be used to predict the weights and the MSE over time. Monte Carte simulations display good to excellent agreement between the actual behavior and the predictions of the theoretical model.
引用
收藏
页码:747 / 756
页数:10
相关论文
共 16 条
[1]
Benedetto S., 1987, DIGITAL TRANSMISSION
[2]
BERMAN A, 1970, IEEE T COMMUN, V18
[3]
BERSHAD NJ, 1993, IEEE T SIGNAL PROCES, V41, P583
[4]
BERSHAD NJ, 1993, IEEE T SIGNAL PROCES, V41, P573, DOI 10.1109/78.993122
[5]
HETRAKUL P, 1976, IEEE T COMMUN, V29, P546
[6]
MULTILAYER FEEDFORWARD NETWORKS ARE UNIVERSAL APPROXIMATORS [J].
HORNIK, K ;
STINCHCOMBE, M ;
WHITE, H .
NEURAL NETWORKS, 1989, 2 (05) :359-366
[7]
IBNKAHLA M, 1996, THESIS NATL POLYTECH
[8]
IBNKAHLA M, IN PRESS IEEE T COMM
[9]
ANALYSIS AND COMPENSATION OF BANDPASS NONLINEARITIES FOR COMMUNICATIONS [J].
KAYE, AR ;
ERIC, MJ ;
GEORGE, DA .
IEEE TRANSACTIONS ON COMMUNICATIONS, 1972, CO20 (05) :965-&
[10]
LIPPMANN RP, 1987, IEEE ASSP MAGAZI APR, P4