Title :
Levenberg-Marquardt training for modular networks
Author :
Fun, Meng-Hock ; Hagan, Martin T.
Author_Institution :
Sch. of Electr. & Comput. Eng., Oklahoma State Univ., Stillwater, OK, USA
Abstract :
The modular neural network has been shown to be an effective alternative to multilayer feedforward networks, especially for implementing functions with sharp changes. This paper describes a new method for training modular networks, based on the Levenberg-Marquardt algorithm for nonlinear least squares. The algorithm is tested on several function approximation problems, and the performance is compared with standard steepest ascent and the Rprop algorithm
Keywords :
Hessian matrices; function approximation; learning (artificial intelligence); least squares approximations; neural nets; optimisation; performance index; Hessian matrix; Levenberg-Marquardt algorithm; function approximation; learning; modular neural network; nonlinear least squares; optimisation; performance index; Approximation algorithms; Function approximation; Least squares approximation; Least squares methods; Multi-layer neural network; Neural networks; Nonhomogeneous media; Performance analysis; Testing; Transfer functions;
Conference_Titel :
Neural Networks, 1996., IEEE International Conference on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-3210-5
DOI :
10.1109/ICNN.1996.548938
