The adaptive learning algorithm for "extended neo-fuzzy neuron" is proposed, that has both following and smoothing properties and allows to solve problems of prediction, filtering and smoothing of non-stationary disturbed stochastic and chaotic signals.
