Learning in fuzzy neural networks

In our fuzzy neural networks, fuzzy weights and fuzzy operations are used for training crisp and fuzzy data. Theoretical studies of fuzzy networks where triangular fuzzy numbers are used, show that the output behaviour of these networks can be estimated for arbitrary input data. To make use of these properties we present two learning algorithms for our networks. We implemented and tested them and found our theoretical observations confirmed. The trained network has the capability of generalizing trained informations well