In this paper, we will try to shed light on the usefulness of neural networks by describing an application which combines connectionism and ruled-based systems. In the present fuzzy ruled production systems, propagation of uncertainty coefficients is carried out by means of computational formulae stemming from mathematical models of fuzzy reasoning. But the use of a formula provided by a general abstact model, and not intimately related to the application, can lead us to a fuzzy procedure not reflecting the fuzzy reasoning of the human expert. The connectionist approach proposed here solves this problem of fuzzy inference. An uncertainty propagation rule specific to the application domain is determined by learning from examples of fuzzy inferences.
[1]
Teuvo Kohonen,et al.
An introduction to neural computing
,
1988,
Neural Networks.
[2]
Yann LeCun,et al.
Modeles connexionnistes de l'apprentissage
,
1987
.
[3]
Widrow,et al.
DARPA Neural Network Stdy
,
1988
.
[4]
W S McCulloch,et al.
A logical calculus of the ideas immanent in nervous activity
,
1990,
The Philosophy of Artificial Intelligence.
[5]
Edward H. Shortliffe,et al.
Computer-based medical consultations, MYCIN
,
1976
.
[6]
W. Pitts,et al.
A Logical Calculus of the Ideas Immanent in Nervous Activity (1943)
,
2021,
Ideas That Created the Future.