Can more robust algorithms be designed to improve the discovery of building blocks? This paper introduces cohort genetic algorithms (cGA's) as a new class of genetic algorithms aimed at enhancing the exploration of search spaces and the exploitation of building blocks in problem-solving. The research presents a novel class of test functions, hyperplane-defined functions (hdf's), designed to evaluate the capabilities of cGA's. The hdf's allow for tracing the origin of each performance advance while resisting reverse engineering, ensuring a fair assessment. The authors conduct experiments to compare the performance of cGA's with traditional genetic algorithms on these hdf's. These capabilities enhance the exploration of search spaces and the exploitation of building blocks already found. The findings suggest that cGA's offer a more robust approach to genetic algorithms, paving the way for advances in various fields where identifying and utilizing building blocks are essential.
Published in Evolutionary Computation, this paper fits squarely within the journal's scope. Evolutionary Computation is focused on exploring the development of artificial intelligence and machine learning through computational and evolutionary processes. The paper's focus on genetic algorithms and their improvement aligns perfectly with the journal's focus.