Title :
Training recurrent networks using Hessian information
Author :
Coelho, Pedro Henrique Gouvêa
Author_Institution :
UFMA, Sao Luis, Brazil
Abstract :
New training algorithms for fully recurrent neural networks are presented. They are based on Hessian matrices estimates. Simulation results show that the algorithms yields satisfactory results
Keywords :
Hessian matrices; estimation theory; learning (artificial intelligence); recurrent neural nets; Hessian information; Hessian matrices estimates; algorithms; fully recurrent neural networks; training algorithms; Approximation algorithms; Backpropagation algorithms; Computational complexity; Computer architecture; Computer networks; Cost function; Electronic mail; Neural networks; Neurons; Recurrent neural networks;
Conference_Titel :
Information Theory, 1995. Proceedings., 1995 IEEE International Symposium on
Conference_Location :
Whistler, BC
Print_ISBN :
0-7803-2453-6
DOI :
10.1109/ISIT.1995.531521
