A Fuzzified Neural Fuzzy Inference Network that Learns from Linguistic Information

A Fuzzified Takagi-Sugeno-Kang (TSK)-type Neural Fuzzy Inference Network (FTNFIN) for handling linguistic information is proposed in this paper. The inputs and outputs of FTNFIN may be fuzzy numbers with any shapes. The α-cut technique is used in input fuzzification and consequent part computation, which enables the network to handle linguistic information. There are no rules in FTNFIN initially since they are constructed on-line by concurrent structure and parameter learning. The network has been applied to the learning of fuzzy input and output data, and good simulation results are achieved.