Reconstructing the shifting balance theory in a GA: taking Sewall Wright seriously

Abstract

We attempt to reconstruct Sewall Wright's (1932) shifting balance theory in order to address some of the major criticisms leveled against it. The resulting abstract process is applied to the GA forming the shifting balance genetic algorithm (SBGA), which is shown to behave as Wright intended. For example, the SBGA avoids local optima through a shifting balance between subpopulations, as is demonstrated in an experiment. The experiment also shows that the SBGA outperforms the classical GA in both stationary and changing environments.