Title :
Recursive neural networks with high capacity
Author :
Chen, Chang-Jiu ; Cheung, John Y.
Author_Institution :
Sch. of Comput. Sci., Oklahoma Univ., Norman, OK, USA
Abstract :
√3n-1 is derived as the lower bound of maximum capacity in n-neuron recursive neural networks. It is shown that if n→∞, the number of stable vectors of (n +1)-neuron net is two times that of n-neuron net and the number of stable vectors of n-neuron net is C2n with 0<C<1. To obtain these results, the SOR method proposed by Oh and Kothari is employed
Keywords :
neural nets; relaxation theory; stability; SOR method; maximum capacity; n-neuron recursive neural networks; stable vectors; successive over-relaxation; Associative memory; CADCAM; Computer aided manufacturing; Computer science; Equations; Geometry; Hebbian theory; Hypercubes; Neural networks; Neurons;
Conference_Titel :
Neural Networks, 1993., IEEE International Conference on
Conference_Location :
San Francisco, CA
Print_ISBN :
0-7803-0999-5
DOI :
10.1109/ICNN.1993.298601
