Abstract: In this paper a new double-wavelet neuron architecture obtained by modification of standard wavelet neuron, and its learning algorithm are proposed. The offered architecture allows to improve the approximation properties of wavelet neuron. Double-wavelet neuron and its learning algorithm are examined for forecasting nonstationary chaotic time series.

Keywords: wavelet, double-wavelet neuron, recurrent learning algorithm, forecasting, emulation, analytical activation function.

ACM Classification Keywords: I.2.6 Learning – Connectionism and neural nets

Link:

DOUBLE-WAVELET NEURON BASED ON ANALYTICAL ACTIVATION FUNCTIONS

Yevgeniy Bodyanskiy, Nataliya Lamonova, Olena Vynokurova

http://www.foibg.com/ijita/vol14/ijita14-3-p14.pdf