SOLVING NON-LINEAR OPTIMIZATION PROBLEMS USING PARALLEL GENETIC ALGORITHM

In last few decades, a growing interest in the domain of evolutionary algorithms has been observed due to its performance in discover the optimal solutions for the complex problems. The Genetic Algorithm (GA) is one of the most used evolutionary algorithms that attract the researchers' interests in many fields such as the physics and mathematics. GA can provide optimal solution for the problems of complex environments i.e. polarize environments. Like other evolutionary algorithms, the execution time of GA is relatively long, whereby the optimization processes could consume many hours. This study aims to improve the optimization accuracy and reduce the execution time of traditional GA. The parallel GA is proposed to conduct the optimization processes through distributed machines or processors (network of processes). The original complex problem is divided into sub-small areas, and each sub area is optimized by sequential GA that applied in each processor in the network. The final solution is collected from all processors in the network in order to decide the best final solution. To evaluate the proposed parallel genetic algorithm, two complex problems are optimized; De Jong's and Ackeleys path functions. For testing purpose, a network of four processors is constructed using MATLAB toolbox distributed system toolbox. The significances results show that proposed parallel GA give better results with less execution time than the traditional or sequential GA. The contribution of this study is the segmentation of large and complex area into small area and optimizes the solutions of the small areas using network of processors. This approach simplifies the optimized problems, reduce the execution time, and give better chances to discover the optimal solutions.