An Asymmetry Subsethood-Based Neural Fuzzy Network

This paper proposes a novel asymmetric subsethood-based neural fuzzy network (ASNFN) that identifies and controls nonlinear dynamic systems. ASNFN has the flexibility to handle both numeric and linguistic inputs. The numeric inputs in ASNFN are fuzzified by input nodes as tunable feature fuzzifiers. Connections in ASNFN are represented by Pseudo-Gaussian fuzzy sets which provide the neural fuzzy network with higher flexibility and which get more accurate optimization. An on-line self-constructing learning algorithm that is constructed and implemented in ASNFN consists of structural learning and parametric learning, and would create adaptive fuzzy logic rules. Computer simulations illustrate the performance and capability of the proposed model in identifying a dynamic system, in Iris data classification, and in backing to park the truck.