Title :
A superior error function for training neural networks
Author :
Kalman, Barry L. ; Kwasny, Stan C.
Author_Institution :
Center for Intelligent Comput. Syst., Washington Univ., St. Louis, MO, USA
Abstract :
The authors present an error function `kerr´ which does not have the objectionable properties exhibited by the error function `ferr´ usually used to train neural networks. When combined with the conjugate gradient method, `kerr´ shows dramatic speedups over `ferr´ and back-propagation. Ferr is a sum of squares; in kerr each square is divided by a factor
Keywords :
errors; learning systems; neural nets; conjugate gradient method; error function; ferr; kerr; modified sum of squares; training neural networks; Computer networks; Feedforward systems; Gradient methods; Intelligent networks; Intelligent systems; Kalman filters; Minimization methods; Neural networks; Terminology; Waste materials;
Conference_Titel :
Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on
Conference_Location :
Seattle, WA
Print_ISBN :
0-7803-0164-1
DOI :
10.1109/IJCNN.1991.155311
