A multi-objective multi-factorial evolutionary algorithm with reference-point-based approach

In recent years, multi-task optimization is one of the emerging topics among evolutionary computation researchers. Multi-Factorial Evolutionary Algorithm (MFEA) is developed based on that individuals, from various cultures, exchange their underlying similarities to improve the convergence characteristic. However, in terms of Multi-Objective Multi-Factorial Optimization (MOMFO), current algorithms employing nondominated front ranking and crowding distance still meet difficulties when the number of objective functions arises. In this paper, we propose a Muli-Objective Multi-Factorial Evolutionary Algorithm (MO-MFEA) with reference-point-based approach to improve the multitasking framework. Rather than using crowding distance to compute individual ranking in the context of MOMFO, we employ a set of reference points to determine the diversity of current population. On the other hand, we improve the guided method that automatically adapt the Random Mating Probability (RMP) in order to exploit shared knowledge among high similar task. Further improvement on genetic operators with JADE crossover and NSLS. The conducted experiments demonstrate our approach performs better than the baseline results.