SIMILARITIES AND DISTINCTIONS IN SAMPLING STRATEGIES FOR GENETIC ALGORITHMS

Following on from a recent report, which presented stochastic models f or two classes of Genetic Algorithms (GAs), we present two results whi ch have important implications with respect to the theoretical basis o f these methods. The first result we present concerns the technique of lumping, and we show how this technique can be used to transform the searching process of a class of GAs. Based on this transformation, our second result concerns the direct comparison of the two main GAs used today, and provides the conditions under which these two GAs are fund amentally distinct search algorithms. A novel role is played by the co nvergent populations in the derivation of these conditions.