MAXIMUM-ENTROPY COMPLEXITY MEASURES

Publisher Website
Google Scholar
Abstract
A maximum-entropy approach to the classification of system complexity is proposed which accounts for assumed probabilistic knowledge concerning the system observables. The approach introduces the concept of conjugate coefficients of complexity and μi-transition points as a generalization of the critical point introduced by Ferdinand for a special case. The results developed include the conditional nature of such a classification method for the general case of an arbitrary number of system observables. The role of maximum-entropy methods to generate system complexity measures is viewed as having principal value in generating the prior, or canonical disiribution ofsystem defects for a given state of knowledge concerning the system observables and as a way to formalize the observer-dependent property of system complexity measures discussed earlier by Ashby.

This publication has 2 references indexed in Scilit: