Lyapunov Exponents as a Nonparametric Diagnostic for Stability Analysis.

The common observation made in the empirical nonlinear dynamics literature is the constraints imposed by the availability of a limited number of observations in the implementation of the existing algorithms of Lyapunov exponents. The algorithm discussed here can estimate all "n" Lyapunov exponents of an unknown "n"-dimensional dynamical system accurately with limited number of observations. This makes the algorithm attractive for applications to economic as well as financial time-series data. The implementation of the algorithm is carried out by multilayer feed forward networks which are capable of approximating any function and its derivatives to any degree of accuracy. Copyright 1992 by John Wiley & Sons, Ltd.