A cost benefit operator for efficient multi level genetic algorithm searches

In this paper we present a novel cost benefit operator that assists multi level genetic algorithm searches. Through the use of the cost benefit operator, it is possible to dynamically constrain the search of the base level genetic algorithm, to suit the user's requirements. Initially we review meta-evolutionary (multi-level genetic algorithm) approaches. We note that the current literature has abundant studies on meta-evolutionary GAs. However these approaches have not identified an efficient approach to termination of base GA search or a means to balance practical consideration such as quality of solution and the expense of computation. Our quality time tradeoff operator (QTT) is user defined, and acts as a base level termination operator and also provides a fitness value for the meta-level GA. In this manner the amount of computation time spent on less encouraging configurations can be specified by the user. Our approach has been applied to a computationally intensive test problem which evaluates a large set of configuration settings for the base GAs. This approach should be applicable across a wide range of practical problems (e.g. routing, logistic and biomedical applications).