thus accelerating distributed training across network nodes. Experimental results demonstrate that the proposed algorithm effectively accelerates large-scale model training tasks.

Abstract: The performance of traditional multiobj ective evolutionary algorithms (MOEAs) often deteriorates rapidly when using them to solve large-scale multiobjective optimization problems (LMOPs).