Title :
Predicting contradictions in the storage process of diluted recurrent Boolean neural networks
Author :
De Padua Braga, Antonio
Author_Institution :
Dept. of Electr. & Electron. Eng., Imperial Coll. of Sci., Technol. & Med., London
fDate :
1/6/1994 12:00:00 AM
Abstract :
Contradictions among patterns cause degradation in the storage capacity of diluted recurrent Boolean neural networks. A model is presented for estimating the likelihood of the occurrence of contradictions, which allows an estimation of the storage capacity of the network. The theoretical model fits very well with computer simulations
Keywords :
Boolean functions; content-addressable storage; neural nets; probability; diluted recurrent Boolean neural networks; model; pattern contradictions; storage capacity; storage process;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19940020
