Some results for chaotic times series prediction using Clifford neural networks

In this work, we consider the extensions of the back error propagation (BEP) learning rule for multilayer perceptrons (MLP) to Clifford algebras and report results on the application of such Clifford neural networks to chaotic time series. Although other researchers have considered similar extensions [(P. Arena et al., 1998), (S. Buchholz and G. Sommer, 2001), (G. Georgiou and C Koutsougeras, 1992), (J .K. Pearson, 1994)] their derivations have been incomplete and not systematically applied. We briefly outline the issues in the derivation of proper learning rules for Clifford neural networks and present a systematic study of a number of multidimensional chaotic time series.