Volterra Type Neo Fuzzy Neuron: A New Model for Identification of Complex and Nonlinear Systems
Noriaki Suetake, Takeshi Yamakawa · IEEJ Transactions on Electronics Information and Systems · 2000
We propose a novel neuron model named Volterra type neo fuzzy neuron (VNFN) model, which is an extension of a conventional neo fuzzy neuron (NFN) model employing the framework of the Volterra series. The VNFN has cross-correlated inputs and is suitable for describing the function which has a correlation among its inputs, such as an exclusive-OR. Furthermore, this neuron model guarantees the convergence to the global minimum in the leaning process as well as a conventional neo fuzzy neuron. The attempt was made to apply this VNFN model to describing the inputs-output characteristics of a function which a conventional NFN model can't describe well. And, it was applied to the modeling and prediction problems of a chaotic time series and a continuous-stirred tank reactor (CSTR) plant.