In neuro-fuzzy (or fuzzy-neural) systems using unconstrained optimization s
chemes like backpropagation, it is not possible to guarantee that the resul
ting membership functions represent human-interpretable linguistic terms. H
owever, one of the most interesting features of fuzzy systems is the insigh
t provided on the linguistic relationship between their variables. This wor
k is devoted to the study of constraints which when used within an optimiza
tion scheme obviate the subjective task of interpreting membership function
s. To achieve this, a comprehensive set of semantic properties that members
hip functions should have is postulated and discussed. Then a set of constr
aints is introduced and shown to be able to fulfil the properties. Implemen
tation issues and one example illustrating the importance of the proposed c
onstraints are included. (C) 1999 Elsevier Science B.V. All rights reserved
.