Most existing multi-objective evolutionary algorithms (MOEAs) have difficulties in approximating the whole Pareto Fronts with complicated geometries. However, the decision maker (DM) may only be interested in a small portion of the front, referred to as the region of interest (ROI). Bearing this in mind, this paper develops a preference-based MOEA with local Pareto Front (PF) estimation to address the above issues. We first modify the r-dominance relation, which is not affected by the position of the user reference point. The modified non-r-dominated solutions are then used as training data during the optimization process to model the PF of the ROI. The estimated local front can be used to drive the search toward the preferred PF region. Experimental results demonstrate the effectiveness of our proposed algorithm on a variety of benchmark problems with different kinds of PFs.
Paper
Full text
A Preference-Based Multi-Objective Evolutionary Algorithm with Local Pareto Front Modeling
Semantic Scholar · Computer Science · 2023
Abstract
Most existing multi-objective evolutionary algorithms (MOEAs) have difficulties in approximating the whole Pareto Fronts with complicated geometries. However, the decision maker (DM) may only be interested in a small portion of the front, referred to as the region of interest (ROI). Bearing this in mind, this paper develops a preference-based MOEA with local Pareto Front (PF) estimation to address the above issues. We first modify the r-dominance relation, which is not affected by the position of the user reference point. The modified non-r-dominated solutions are then used as training data during the optimization process to model the PF of the ROI. The estimated local front can be used to drive the search toward the preferred PF region. Experimental results demonstrate the effectiveness of our proposed algorithm on a variety of benchmark problems with different kinds of PFs.