Enhancing MOEA/D with information feedback models for large-scale many-objective optimization

A multi-objective evolutionary algorithm based on decomposition (MOEA/D) is a classic decomposition-based multi-objective optimization algorithm. In the standard MOEA/D algorithm, the update process of individuals is a forward search process without using the information of previous individuals. However, there is a lot of useful information in the previous iteration. Information Feedback Models (IFM) is a new strategy which can incorporate the information from previous iteration into the updating process. Therefore, this paper proposes a MOEA/D algorithm based on information feedback model, called MOEA/D-IFM. According to the different information feedback models, this paper proposes six variants of MOEA/D, and these algorithms can be divided into two categories according to the way of selecting individuals whether it is random or fixed. At the same time, a new selection strategy has been introduced to further improve the performance of MOEA/DIFM. The experiments were carried out in four aspects. MOEA/D-IFM were compared with other state-of-the-art multi-objective evolutionary algorithms using CEC 2018 problems in two aspects. The best one of the six improved algorithms was chosen to test on large-scale many-objective problems. In addition, we also use MOEA/D-IFM to solve multi-objective backpack problems. (C) 2020 Elsevier Inc. All rights reserved.