Improved measures of robustness and evolvability for evolutionary systems

Abstract

This research reviews existing definitions of evolvability and robustness and introduces new baseline measures that addresses the drawbacks of the widely used definitions. These new measures are applied to a variety of systems and problems to demonstrate their versatility and ease of use, including a new and simple modeling approach (BNK). This new modelling approach is similar to NK System but with the advantage of providing both a genotype and a phenotype. It is used to demonstrate the ease of evolving oscillatory circuit using different system configurations and their robustness to state perturbation, while also reviewing the relationship between the complexity of the system and the resulting multi-modality of its fitness surface. Furthermore, the existing measures are compared with the proposed measures using RNA sequence folding to assess their respective ability to capture evolvability and robustness. In addition, a demonstration of application of the new measures to linear variants of Genetic Programming is also provided as concrete evidence of their ease of use.

Date	19 Dec 2023
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Patrick Cardinal (Supervisor) & Nawwaf Kharma (Co-supervisor)