Information content and maximum entropy of compartmental systems in equilibrium

TL;DR

This paper introduces entropy measures for compartmental systems modeled as Markov chains, enabling better model selection through maximum entropy principles by quantifying the uncertainty of particle paths and system dynamics.

Contribution

It provides explicit formulas for path entropy and entropy rates in equilibrium compartmental systems, linking them to Shannon entropy and aiding model selection.

Findings

Derived explicit formulas for path entropy and entropy rates.

Demonstrated how entropy measures can resolve model equifinality.

Applied entropy measures to improve model selection in compartmental systems.

Abstract

Although compartmental dynamical systems are used in many different areas of science, model selection based on the maximum entropy principle (MaxEnt) is challenging because of the lack of methods for quantifying the entropy for this type of systems. Here, we take advantage of the interpretation of compartmental systems as continuous-time Markov chains to obtain entropy measures that quantify model information content. In particular, we quantify the uncertainty of a single particle's path as it travels through the system as described by path entropy and entropy rates. Path entropy measures the uncertainty of the entire path of a traveling particle from its entry into the system until its exit, whereas entropy rates measure the average uncertainty of the instantaneous future of a particle while it is in the system. We derive explicit formulas for these two types of entropy for compartmental systems in equilibrium based on Shannon information entropy and show how they can be used to solve equifinality problems in the process of model selection by means of MaxEnt.