A Memetic Algorithm with Non Gradient-Based Local Search Assisted by a Meta-Model


The development of multi-objective evolutionary algorithms (MOEAs) assisted by meta-models has increased in the last few years. However; the use of local search engines assisted by meta-models for multi-objective optimization has been less common in the specialized literature. In this paper, we propose the use of a local search mechanism which is assisted by a meta-model based on support vector machines. The local search mechanism adopts a free-derivative mathematical programming technique and consists of two main phases: the first generates approximations of the Pareto optimal set. Such solutions are obtained by solving a set of aggregating functions which are defined by different weighted vectors. The second phase generates new solutions departing from those obtained during the first phase. The solutions found by the local search mechanism are incorporated into the evolutionary process of our MOEA. Our experiments show that our proposed approach can produce good quality results with a budget of only 1;000 fitness function evaluations in test problems having between 10 and 30 decision variables.