ABSTRACT
Genetic algorithms (GAs) are a class of evolutionary algorithms first proposed and analyzed by John Holland (1975). There are three features which distinguish GAs, as first proposed by Holland, from other evolutionary algorithms: (i) the representation used-bitstrings (Chapter 15); (ii) the method of selectionproportional selection (Chapter 23) ; and (iii) the primary method of producing variations-crossover (Chapter 33). Of these three features, however, it is the emphasis placed on crossover which makes GAs distinctive. Many subsequent GA implementations have adopted alternative methods of selection, and many have abandoned bitstring representations for other representations more amenable to the problems being tackled. Although many alternative methods of crossover have been proposed, in almost every case these variants are inspired by the spirit which underlies Holland’s original analysis of GA behavior in terms of the processing of schemata or building blocks. It should be pointed out, however, that the evolution strategy paradigm (Chapter 9) has added crossover to its repertoire, so that the distinction between classes of evolutionary algorithms has become blurred (Ba¨ck et al 1991).
