Abstract:
Neuro-fuzzy systems are soft computing methods utilizing artificial neural networks and fuzzy systems. Various connectionist architectures of neuro-fuzzy systems can be considered. When systems of this kind solve a problem, they perform according to fuzzy IF-THEN rules, which constitute a knowledge base. The knowledge acquisition realized by intelligent systems is very important from application point of view. Different methods can be employed. A new, perception-based, approach that imitates the human way of knowledge acquisition is proposed. When applied as classifiers, the systems that utilize this knowledge, perform without misclassifications. The systems can incorporate type 2 fuzzy sets and other soft computing methods, e.g. probability theory, evolutionary algorithms, as well as rough sets. The hybrid approach can be considered in the framework of multi-expert systems. The perception-based approach refers to the computing with words introduced by Prof. Zadeh.