A simple neural model for fuzzy reasoning
J.A.B. Tome · 2002
A very simple neural architecture for fuzzy reasoning is presented. It is shown that fuzzy rules may be implemented with such nets. The net is layered and the concept of variables and predicates may be associated with areas in those layers. It is the density of activated neurons which defines the membership grades. Fuzzy logic operations are induced in a natural way by the random connections of the neurons from layer to layer. The layered structure of the model, its simplicity and the randomness of the connections makes this model adequate for representing natural systems.>