Hardware Realization of Fuzzy Neural Networks

It is needless to say that a panoply of real-world applications of fuzzy sets call for a variety of systems realizing fuzzy computation. Concurrently, it is highly desirable to develop some universal computing modules that may be easily customized to meet required hardware — software specifications. This concept is not that new, as in two-valued logic we can easily encounter various categories of configurable and programmable devices (PLDs), ranging from PAL’s to FPGA’s. These devices are standardized general purpose logic units, which may be configured to perform specific functions. To follow a similar avenue, it is indispensable to identify a few generic processing modules that are complete enough, when considered from a functional point of view, for general computations with fuzzy linguistic variables. A family of logic-based neurons [1][2], emerges as a collection of processing operations whose role is to model logic-oriented processing dominant in the theory of fuzzy sets. These configurable architectures arising within this framework can directly cope with the topology of the problem at hand.