Multisurrogate-Assisted Multitasking Particle Swarm Optimization for Expensive Multimodal Problems.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Expensive multimodal optimization problems (EMMOPs) are prevalent in real-world applications.
• Surrogate-assisted evolutionary algorithms (SAEAs) face challenges in surrogate model selection and multimodal discovery.
• Diverse optimization scenarios necessitate adaptive surrogate model strategies.

Purpose of the Study:

• To develop a cost-effective algorithm for finding multiple optimal solutions in EMMOPs.
• To integrate multiple surrogate models and multitasking optimization for enhanced performance.
• To improve the balance between computational cost and solution accuracy in optimization.

Main Methods:

• A multisurrogate-assisted multitasking particle swarm optimization algorithm is proposed.
• EMMOPs are transformed into multitasking problems using various surrogate models.
• A multitasking niche particle swarm algorithm is designed, incorporating a skill factor and clustering-based surrogate model management strategy.
• An adaptive local search strategy based on the trust region is employed.

Main Results:

• The proposed algorithm effectively seeks multiple optimal solutions for EMMOPs at a low computational cost.
• Experimental results on 19 benchmark functions and a building energy conservation problem demonstrate superior performance.
• The algorithm achieves competitive results compared to state-of-the-art SAEAs and multimodal EAs.

Conclusions:

• The developed algorithm offers an efficient approach to tackle EMMOPs by leveraging multitasking and multisurrogate strategies.
• The surrogate model management and adaptive local search enhance the discovery of multiple optima.
• This work provides a valuable tool for addressing complex optimization challenges in various domains.