Novel multi-objective optimization algorithm

Many multi-objective evolutionary algorithms (MOEAs) can converge to the Pareto optimal front and work well on two or three objectives, but they deteriorate when faced with many-objective problems. Indicator-based MOEAs, which adopt various indicators to evaluate the fitness values (instead of the Pareto-dominance relation to select candidate solutions), have been regarded as promising schemes that yield more satisfactory results than well-known algorithms, such as non-dominated sorting genetic algorithm (NSGA-II) and strength Pareto evolutionary algorithm (SPEA2). However, they can suffer from having a slow convergence speed. This paper proposes a new indicator-based multi-objective optimization algorithm, namely, the multi-objective shuffled frog leaping algorithm based on the e indicator (є-MOSFLA). This algorithm adopts a memetic meta-heuristic, namely, the SFLA, which is characterized by the powerful capability of global search and quick convergence as an evolutionary strategy and a simple and effective є-indicator as a fitness assignment scheme to conduct the search procedure. Experimental results, in comparison with other representative indicator-based MOEAs and traditional Pareto-based MOEAs on several standard test problems with up to 50 objectives, show that є-MOSFLA is the best algorithm for solving many-objective optimization problems in terms of the solution quality as well as the speed of convergence.