A class of global optimization algorithms that begin with a loose analogy between optimization and a biological system composed of a relatively few organisms that react in a relatively complex way. Algorithms try to evolve a population of trial members in a way mimicking biological evolution. Points in the domain are called ‘‘models’’ and each model has a ‘‘fitness’’ associated with it; the goal is to find the most fit of possible models. A genetic algorithm is a set of operations that we apply to a population of models to produce a new population whose average fitness exceeds that of its predecessors. The characteristics of models are specified by ‘‘chromosome strings.’’ One type of genetic algorithm selects parents randomly but weighted by their fitness (selection); the chromosomes for the ‘‘child’’ are somewhat randomly selected from the two parents (crossover). The child then joins the population and the least fit member of the population (which might be the child) is eliminated. At random times a ‘‘mutation,’’ a random change in a member’s chromosomes, occurs; this permits introducing into the species chromosome elements not present in the original population. See Smith et al. (1992) and Stoffa and Sen (1991).