A new adaptive approach, which applies the concept of relative importance of information to the genetic algorithm, is developed. Selection, crossover and mutation are parameter-free in the sense that the problem at a particular stage of evolution chooses the parameters automatically. Both fitness and allele statistics are used as information measures to guide the search. Fitness, taking the role as it is in the conventional genetic algorithm, indicates how well a solution can solve the problem. Allele statistics, a new tool to the genetic algorithm, reveals hidden structural information of the solutions. Selecting fit chromosomes to survive to the next generation, choosing unfit chromosomes to mutate, and selection of good parents for crossover are based on the fitness. The allele stati...[ Read more ]

A new adaptive approach, which applies the concept of relative importance of information to the genetic algorithm, is developed. Selection, crossover and mutation are parameter-free in the sense that the problem at a particular stage of evolution chooses the parameters automatically. Both fitness and allele statistics are used as information measures to guide the search. Fitness, taking the role as it is in the conventional genetic algorithm, indicates how well a solution can solve the problem. Allele statistics, a new tool to the genetic algorithm, reveals hidden structural information of the solutions. Selecting fit chromosomes to survive to the next generation, choosing unfit chromosomes to mutate, and selection of good parents for crossover are based on the fitness. The allele statistics are used to choose loci to undergo mutation and to be swapped between chromosomes in crossover. The Hamming distance is also employed in the crossover to choose suitable parents. The allele statistics are useful in solving problems for which the loci of the encoded chromosomes exhibit the relative importance of structural information, such as the knapsack problem.