Genetic Local Search Algorithm with Self-Adaptive Population Resizing for Solving Global Optimization Problems

Abstract

In the past decades, many types of nature inspired optimization algorithms have been proposed to solve unconstrained global optimization problems. In this paper, a new hybrid algorithm is presented for solving the nonlinear unconstrained global optimization problems by combining the genetic algorithm (GA) and local search algorithm, which increase the capability of the algorithm to perform wide exploration and deep exploitation. The proposed algorithm is called a Genetic Local Search Algorithm with Self-Adaptive Population Resizing (GLSASAPR). GLSASAPR employs a self-adaptive population resizing mechanism in order to change the population size NP during the evolutionary process. Moreover, a new termination criterion has been applied in GLSASAPR, which is called population vector (PV) in order to terminate the search instead of running the algorithm without any enhancement of the objective function values. GLSASAPR has been compared with eight relevant genetic algorithms on fifteen benchmark functions. The numerical results show that the proposed algorithm achieves good performance and it is less expensive and cheaper than the other algorithms.