Multi-objective multifactorial evolutionary algorithm enhanced with the weighting helper-task

Recently, transfer learning has received more and more attention in the field of computational intelligence. The multi-task paradigm is a recent research hotspot. Among them, multi-objective multitasking optimization aims to optimize multiple multi-objective optimization problems simultaneously. The first evolutionary algorithm for multi-objective multitasking optimization is multi-objective multifactorial algorithm (MO-MFEA). However, MO-MFEA has slow convergence due to irrelevance or weakly relevance among tasks. To deal with this issue, we introduce an additional helper-task, i.e., a weight sum of component tasks, into MO-MFEA to improve the effectiveness of inter-task knowledge transfer. Experimental results on a set of benchmark problems have validated the effectiveness and efficiency of the proposed method as compared with MOMFEA and NSGA-II.