Heterogeneous parallel algorithms to solve epistatic problems

We propose parallel heterogeneous metaheuristics (PHM) to solve a kind of epistatic problem (NK-Landscape). The main feature of our heterogeneous algorithms is the utilization of multiple search threads using different configurations to guide the search process. We propose an operator-based PHM, where each search thread uses a different combination of recombination and mutation operators. We compare the performance of our heterogeneous proposal against an homogeneous algorithm (multiple threads with the same parameter configuration) in a numerical and real time ways. Our experiments show that the heterogeneity could help to design powerful and robust optimization algorithms on high dimensional landscapes with an additional reduction in execution times.