Abstract
In this work, we propose a framework to accelerate the computational efficiency of evolutionary algorithms on largescale multi-objective optimization. The main idea is to track the Pareto optimal set directly via decision space reconstruction. To begin with, the algorithm obtains a set of reference directions in the decision space and associates them with a set of weight variables for locating the Pareto optimal set. Afterwards, the decision space is reconstructed by taking the weight variables and their corresponding solutions as the input and output of the reconstructed optimization problem, respectively. Thanks to the low dimensionality of the weight variables, a set of quasi-optimal solutions can be obtained efficiently. Finally, a multi-objective evolutionary algorithm is used to spread the quasi-optimal solutions over the approximate Pareto optimal front uniformly. Experiments have been conducted on a variety of large-scale problems with 2 or 3 objectives and up to 1000 decision variables. Four different types of well-known algorithms are embedded into the proposed framework and compared with their original versions, respectively. Furthermore, the proposed framework has been compared with two state-of-the-art algorithms for largescale multi-objective optimization. Experimental results have demonstrated the significant improvement benefited from the framework in terms of its performance and computational efficiency in large-scale multi-objective optimization.