Local Lyapunov Exponents: Looking Closely at Chaos
Journal of the Royal Statistical Society: Series B (Statistical Methodology) · 1992 · Volume 54 · Issue 2
DOI: 10.1111/j.2517-6161.1992.tb01886.x
Abstract
When a deterministic mechanism gives rise to an erratic time series, then under certain conditions the series is said to exhibit chaos. The catalogue of statistical methods for the analysis of time series is being augmented currently by methods for the detection and quantification of deterministic chaos. A hallmark of chaos is the tendency of nearby trajectories to diverge in the short term. A measure of the average rate of exponential divergence exhibited by a chaotic system is given by the Lyapunov exponents of that system; the values of such exponents can suggest the presence of chaos. In this paper, we give a brief exposition of a finite time version of the Lyapunov exponent which is state dependent, and to which we attach the epithet `local'. The local Lyapunov exponent is shown to be able to detect non-linearity in some cases, and to have other qualitative features. As defined, its success is limited, but this does not preclude its usefulness as a diagnostic tool.
Keywords
chaos, lyapunov exponent, observation noise, sensitivity to initial conditions, system noise, trajectory
Enable JavaScript for citation tools and additional paper interactions.