Title :
On the relative time complexities of standard and conjugate gradient backpropagation
Author :
Stone, James V. ; Lister, Raymond
Author_Institution :
Sch. of Cognitive & Comput. Sci., Sussex Univ., Brighton, UK
fDate :
27 Jun-2 Jul 1994
Abstract :
It is well known that conjugate gradient backpropagation is much faster than standard backpropagation on many given problem instances. However, it has not been established that the conjugate gradient form has a lower time complexity than the standard algorithm. We describe an empirical estimation of the algorithms´ respective time complexities, on N2N encoder problems. The algorithms were found to have near equal median time complexities, approximately O(N4)
Keywords :
backpropagation; computational complexity; conjugate gradient methods; neural nets; N2N encoder; conjugate gradient backpropagation; neural nets; relative time complexities; standard backpropagation; Algorithm design and analysis; Backpropagation; Biology computing; Computer science; Equations; Error correction; Optimization methods; Pathology; Polynomials; Size control;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374143