Scholarlink

<p>This paper introduces MOMEVO (Multi-Objective Metamodel-based Evolutionary Optimizer), a surrogate-assisted multi-objective evolutionary algorithm developed to address the challenges of computationally expensive optimization tasks commonly encountered in engineering and other domains. One of the core contributions of MOMEVO is its distance-based candidate selection mechanism, promoting global exploration more effectively than traditional criteria. Additionally, it features an iterative search-space reduction framework that adaptively focuses on high-potential regions while maintaining global coverage. MOMEVO’s modular architecture allows dynamic switching between Gaussian Processes and ensemble models, enhancing robustness in challenging landscapes. To explore the solution space effectively, MOMEVO uses evolutionary operators such as crossover and Gaussian mutation. Comprehensive experimental evaluations were conducted on synthetic multi-objective benchmark problems. Results demonstrate that MOMEVO delivers competitive or superior performance relative to established surrogate-assisted algorithms in terms of inverted generational distance (IGD), maximum spread, and computational efficiency. Additionally, results show that MOMEVO is significantly more efficient, completing optimization nearly three times faster than MOBopt and over five times faster than ParEGO and TSEMO. These findings highlight MOMEVO’s robustness, efficiency, and potential for solving real-world multi-objective optimization problems under limited evaluation budgets.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

MOMEVO: a modular surrogate-assisted evolutionary framework for efficient multi-objective optimization

  • 期刊论文

  • 发表日期：

    2026年4月13日

查看全文
  • Rafael Batres,
  • Azul Rosales

摘要

This paper introduces MOMEVO (Multi-Objective Metamodel-based Evolutionary Optimizer), a surrogate-assisted multi-objective evolutionary algorithm developed to address the challenges of computationally expensive optimization tasks commonly encountered in engineering and other domains. One of the core contributions of MOMEVO is its distance-based candidate selection mechanism, promoting global exploration more effectively than traditional criteria. Additionally, it features an iterative search-space reduction framework that adaptively focuses on high-potential regions while maintaining global coverage. MOMEVO’s modular architecture allows dynamic switching between Gaussian Processes and ensemble models, enhancing robustness in challenging landscapes. To explore the solution space effectively, MOMEVO uses evolutionary operators such as crossover and Gaussian mutation. Comprehensive experimental evaluations were conducted on synthetic multi-objective benchmark problems. Results demonstrate that MOMEVO delivers competitive or superior performance relative to established surrogate-assisted algorithms in terms of inverted generational distance (IGD), maximum spread, and computational efficiency. Additionally, results show that MOMEVO is significantly more efficient, completing optimization nearly three times faster than MOBopt and over five times faster than ParEGO and TSEMO. These findings highlight MOMEVO’s robustness, efficiency, and potential for solving real-world multi-objective optimization problems under limited evaluation budgets.