Flexible wavelet-neuro-fuzzy neuron in dynamic data mining tasks
Yevgeniy V. Bodyanskiy, Iryna Pliss, Olena А. Vynokurova · Electronic scientific archive of UrFU (Ural Federal University) · 2013
A new flexible modification of neo-fuzzy neuron (FNFN) and adaptive learning algorithms for the tuning of its all parameters are proposed in the paper. The algorithms are interesting in that they provide on-line tuning of not only the synaptic weights and membership functions parameters, but also forms of these functions, that provide improving approximation properties and allow to avoid the occurrence of ”gaps” in space of inputs. The proposed algorithms have both the tracking and filtering properties, so the FNFN can be effectively used for prediction, filtering and smoothing of non-stationary stochastic and chaotic sequences. A special feature of the proposed approach is its computational simplicity in comparison with known learning procedures for hybrid wavelet-neuro-fuzzy systems of computational intelligence.