.png){kind=icon}
.png){kind=icon}
Return
Composite indicator-guided infilling sampling for expensive multi-objective optimization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2026.102312.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In computationally expensive multi-objective optimization, where the evaluation budget is severely limited, the selection of promising candidate solutions for costly fitness evaluations plays a critical role in accelerating convergence and enhancing algorithmic performance. Nevertheless, devising an optimization strategy that effectively balances convergence, diversity, and distribution remains a challenging task. To address this issue, this paper proposes a composite indicator-based evolutionary algorithm (CI-EMO) for expensive multi-objective optimization. During each generation of the optimization process, CI-EMO explores the solution space based on Gaussian Process model assisted NSGA-III, thereby generating a candidate population. Subsequently, a novel composite performance indicator is introduced to guide the selection of candidates for actual fitness evaluation. This indicator simultaneously accounts for convergence, diversity, and distribution, thereby enhancing the efficiency of identifying promising candidate solutions and significantly boosting algorithmic performance. The proposed composite indicator-based candidate selection strategy is straightforward to implement and computationally lightweight. Component analysis experiments validate the effectiveness of each constituent within the composite performance indicator. Comparative studies conducted on three benchmark test suites and real-world problems demonstrate that the proposed algorithm outperforms five state-of-the-art algorithms for expensive multi-objective optimization.
Keywords:
Expensive multi-objective optimization
Gaussian Process model
NSGA-III
Composite performance indicator
Infilling sampling
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W