Fuzzy models, modular networks, and hybrid learning
- Additional Document Info
- View All
This paper proposes a new approach that can integrate fuzzy logic and neural networks in a "natural" manner. Unlike most existing fuzzy-neural models which usually make use of the structure of feedforward multilayer networks, the proposed model takes advantage of the structure of a kind of modular networks. We show that fuzzy models have a direct correspondence with the modular networks. Based on this correspondence, we develop an efficient hybrid learning scheme which combines an unsupervised learning algorithm (fuzzy-c-means algorithm) and a supervised algorithm (LMS algorithm). The utility of the proposed approach is illustrated using the well-known Zimmermann and Zysno data . © 1996 - Elsevier Science B.V. All rights reserved.
author list (cited authors)