Efficient Metaheuristics in Higher-level Programming Languages

This paper studies the impact of modern programming languages in the design of metaheuristic algorithms. Well-known metaheuristics have been proposed and studied over the last decades, including population-based and evolutionary algorithms like Genetic and Memetic Algorithms, trajectory-based Simulated Annealing, Tabu Search, Variable Neighborhood Search, among others. Despite efforts from community in proposing generic metaheuristic frameworks in low-level languages like C/C++, the growth of developers in higher-level programming languages --- such as Julia and  Python --- demands existing efficient C/C++ tools to be adapted for the languages, so that new users can also benefit from them. This work proposes a Julia version of the OptFrame, a C/C++ combinatorial optimization framework for metaheuristics, comparing performance with other higher-level languages, like Python, considering well-known challenging combinatorial optimization problems. Results indicate that Julia is able to provide similar C/C++ performance, outpacing Python, while being easier to code and explore newer efficient metaheuristic techniques.