Computational Issue of Fuzzy Rule-based System

An innovative soft-computing system is proposed in this paper to assuage the paradox of curse of dimensionality (COD) and yet to preserve the property of completeness. Although the supreme merits of a fuzzy inference system (FIS) are in its simplicity, understandability of fuzzy rules and model-free approach, it suffers with the problem of COD. The COD problem can be happened easily if the number of either input variables or partitions of each input universe increases greatly. This drawback of COD causes serious situation especially for hardware realization that computational resources will be sucked exhaustively. An event-triggering based neuro-fuzzy system (NFS) is presented to alleviate the COD problem and to reduce wasting computational resources. The proposed soft-computing system can save the computational resources effectively without missing the property of completeness. The input to the proposed NFS is considered as an event. Because the incoming input event H(t) decides the position in the input space around which fuzzy sets with membership degree beyond a threshold are detected and are used to construct the fuzzy rules closely to the H(t), the fuzzy rules involved in the knowledge base can be very compact. The event-triggering based knowledge base is effective, in which there are no redundant fuzzy rules and only the needed rules are in the proposed system for the H(t). Moreover, the proposed soft-computing system, whose structure is time-varying and is dependent on the incoming input event, possesses the property of event-tracking structure. The knowledge base of the proposed NFS is triggered off by the event, and only few rules are fired locally around the event. It is suitable for large-scale system operation. An example is demonstrated for the proposed approach.