An A* like algorithm to improve the performance of geneticalgorithms

摘要

Genetic algorithms (GAs) often suffer from difficulties ofconvergence due to lack of guidelines for the selection process.Normally the selection is based on the current fitness of theindividuals, evaluated by a fitness function. However, the presentfitness of an individual need not always indicate its ability to improvefurther. In this work, we propose an A* like evaluation method, whichtakes into account not only the present fitness of the individual, butalso an estimate of its scope for further improvisation. This simpleimprovement to simple GA has produced results comparable to specialisedGA methods in selection problems