Parallelization of crossover operators of a genetic algorithm

Abstract

The study focuses on the optimization of genetic algorithms (GAs) through the parallelization of the crossover operator, inspired by Darwin's evolutionary theory. The objective is to improve the performance and scalability of GAs to address complex problems, such as agricultural optimization. The research followed a structured experimental methodology in five phases: preliminary research, design, implementation, experimentation, and analysis. Python and specialized libraries for parallel processing were used, optimizing the crossover operator through parallelization techniques and GPU acceleration.

Results show that parallelization reduced execution time by 20% and improved solution quality by 25%, particularly in high-complexity problems. The three-point crossover operator proved to be the most efficient option by balancing genetic diversity, performance, and execution time. The use of parallel computational resources was efficient and sustainable, demonstrating its applicability in real-world settings.

Therefore, it can be stated that the parallelization of the crossover operator in GAs improves both computational efficiency and solution quality, establishing itself as a strategic tool for solving large-scale problems across various fields.