Expensive multiobjective immune algorithm using a novel differential evolution in objective space

doi:10.1016/j.eswa.2025.129708

Expensive multiobjective immune algorithm using a novel differential evolution in objective space

2025-09-25

0

PRE

AI

DOI：10.1016/j.eswa.2025.129708

Original

Original request for help

Save

Abstract

En 中文

Generating offspring solutions with strong convergence and diversity is critical when solving expensive multiobjective optimization problems due to the limited number of objective function evaluations. However, existing algorithms produce offspring solutions in the decision space, causing significant uncertainty in obtaining offspring with strong convergence and diversity. To address this issue, we devise a novel differential evolution based on the objective space rather than the decision space, called Differential Objective Evolution (DOE). Specifically, DOE generates objective values with strong convergence and diversity and then maps these values to the decision space to achieve high-quality offspring. Furthermore, we utilize a multiobjective immune algorithm to produce high-quality samples for effectively training the mapping in DOE. When compared with eleven recently proposed algorithms on 105 expensive multiobjective optimization problems, the experiments demonstrate the superiority of our algorithm and the contributions of DOE in both population convergence and diversity.

Expensive multiobjective immune algorithm using a novel differential evolution in objective space

Abstract

Journal

Organization