A Coevolutionary Algorithm with Diffusive Growing Neural Gas Strategy for Many-Objective Optimization Problems with Irregular Pareto Front

Abstract Due to the difficulty in balancing convergence, diversity and irregular Pareto fronts (PFs) handling simultaneously, most existing evolutionary algorithms face significant challenges on many-objective optimization problem with irregular PFs. To address these challenges, this article proposes a coevolutionary algorithm with diffusive growing neural gas (GNG) strategy for many-objective optimization problems with irregular Pareto front, called CoA-DGNG. Specifically, a coevolutionary mechanism is designed to collaborative facilitate both objective optimization and irregular PFs handling. Meanwhile, by utilizing the topology information of PFs, a diffusive growing neural gas node update strategy, named DGNG, is developed to guide the update of adaptive reference vectors. Moreover, a two-fold environmental selection strategy is presented to enhance selection pressure towards promising direction. The proposed algorithm is compared with four state-of-the-art optimizers on 45 test problems with irregular PFs. The experimental results demonstrate the superiority of the proposed CoA-DGNG in comparison to the chosen algorithms.