Enhance the Baldwin effect by strengthening the correlation between genetic operators and learning methods

Abstract

Baldwin effect is one of the popular approaches to combining evolutionary search and learning. However, the benefit of the Baldwin effect can be difficult to obtain in some circumstances. This is due to the lack of understanding of the factors affecting the Baldwin effect. This work illustrates that the Baldwin effect is affected by the correlation between genetic operators and learning methods. Examples are provided to demonstrate how to improve the performance of Baldwinian learning by adjusting the genetic operators towards a stronger correlation with learning.