Statistical versus optimal partitioning for block entropies

Abstract

Purpose

Given a time‐series, what is the best partitioning of the state space in order to obtain reasonable values for the block entropies? The purpose of this paper is to provide a simple answer (an algorithm), although approximative, in connection with symbolic dynamics and statistical properties of 1‐d maps on the interval.

Design/methodology/approach

The logistic map is examined as an archetype of a Complex System with different behaviors, namely: periodicity, order‐to‐chaos period‐doubling transition, weak chaos, parametric intermittent chaos, developed chaos and fully developed chaos. For the logistic map the generating partition is known, and allows comparison with other prescriptions in the literature. The partitioning of the phase space with the easy generated bipartition induced by the mean value of a curve in the plane, gives results in good agreement (roughly up to a 20 per cent difference) with the results of the generating partition, if the trajectory of the system is in parametric intermittent chaos and in developed chaos (DC). In the case of fully developed chaos (FDC), the agreement is perfect.

Findings

The authors confirm that a statistical partitioning is almost equivalent with the exact partitioning for the logistic map.

Originality/value

The paper updates previous results and proposes a better understanding on the partitioning for symbolic dynamics.

Citation

Mistakidis, I., Karamanos, K. and Mistakidis, S. (2013), "Statistical versus optimal partitioning for block entropies", Kybernetes, Vol. 42 No. 1, pp. 35-54. https://doi.org/10.1108/03684921311295466