Redundant binary representations with rigorous trade-off between connectivity and locality

Abstract

We design a general mathematical framework of redundant binary representations that are easy to apply for evolutionary algorithms and are efficient with respect to several measures such as synonymity, connectivity, and locality. The proposed representations are easy to scale for any problem dimension and any level of neutrality, and they can exhibit different values of connectivity and locality.

Keywords

• connectivity
• fitness distance correlation
• locality
• neutral networks
• phenotype mutation rates
• problem hardness
• synonymity