A novel parameter for nonequilibrium analysis in reconstructed state spaces

TL;DR

This paper introduces a new nonequilibrium descriptor based on kernel probabilistic difference in reconstructed state spaces, effectively analyzing chaotic and heartbeat data to quantify physiological complexity loss.

Contribution

It proposes a novel nonequilibrium measure using kernel probabilistic difference, especially with Kullback-Leibler divergence, for analyzing reconstructed state spaces.

Findings

Effectively characterizes complexity loss in heartbeat data.

Successfully distinguishes physiological and pathological cardiac states.

Validated on chaotic series and real-world heartbeat data.

Abstract

Kernel methods are widely used for probability estimation by measuring the distribution of low-passed vector distances in reconstructed state spaces. However, the information conveyed by the vector distances that are greater than the threshold has received little attention. In this paper, we consider the probabilistic difference of the kernel transformation in reconstructed state spaces, and derive a novel nonequilibrium descriptor by measuring the fluctuations of the vector distance with respect to the tolerance. We verify the effectiveness of the proposed kernel probabilistic difference using three chaotic series (logistic, Henon, and Lorenz) and a first-order autoregressive series according to the surrogate theory, and we use the kernel parameter to analyze real-world heartbeat data. In the heartbeat analysis, the kernel probabilistic difference, particularly that based on the Kullback--Leibler divergence, effectively characterizes the physiological complexity loss related to reduced cardiac dynamics in the elderly and diseased heartbeat data. Overall, the kernel probabilistic difference provides a novel method for the quantification of nonequilibria by characterizing the fluctuation theorem in reconstructed state spaces, and enables reliable detection of cardiac physiological and pathological information from heart rates.