In the paper, the fuzzy connection networks related to max-min fuzzy Hopfie
ld networks are introduced to analyze the attractors and attractive cycles
of the max-min networks. The elementary memories are employed to improve fa
ult-tolerance of the networks. Some sufficient and necessary conditions tha
t an attractor is the elementary memory of the networks are shown. Homomorp
hism operators on [0.1]" are utilized to study the transitive trajectories
of the states, attractors and attractive cycles of max-min Hopfield network
s. By the conclusions obtained here, a novel way for designing fuzzy neural
networks with better fault-tolerance may be constructed. (C) 2001 Elsevier
Science B.V. All rights reserved.