Determining the Information Flow of Dynamical Systems from Continuous Probability Distributions

Authors:

Abstract
We present a technique to estimate the information flow in dynamical systems. The information flow characterizes the underlying dynamics which can be stochastic, deterministic chaotic, or deterministic nonchaotic. For chaos our approach offers a method for calculating the Kolmogorov-Sinai (KS) entropy from continuous probability density functions. These density functions are approximated by neural networks. For stochastic processes (noisy chaos, autoregressive models) our approach yields the rate of memory loss of the dynamics.