Tuning a Hybrid Optimization Algorithm by Determining the Modality of the Design Space

In this paper we present an approach for increasing the efficiency of a hybrid Genetic/Sequential Linear Programming algorithm. We introduce two metrics for evaluating the modality of the design space and then use this information to efficiently switch between the Genetic Algorithm and SLP algorithm. The motivation for this study is an effort to reduce the computational expense associated with the use of a Genetic Algorithm by reducing the number of function evaluations needed to find good solutions. In the paper the two metrics used to evaluate the modality of the design space are the variance in fitness of the population of the designs in the Genetic Algorithm and the error associated with fitting a response surface to the designs evaluates by the Genetic Algorithm. The effectiveness of this approach is demonstrated by considering a highly multimodal Genetic Algorithm benchmarking problem.