Localization for solving noisy multi-objective optimization problems.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Multi-objective optimization

Background:

• Evolutionary multi-objective optimization (EMO) algorithms face challenges with noisy environments.
• Noise can degrade algorithm performance and reduce solution robustness.
• Existing global models may struggle to effectively filter noise in complex search spaces.

Purpose of the Study:

• To investigate a novel framework of local models for noisy EMO.
• To enhance the noise-filtering capabilities and robustness of EMO algorithms.
• To compare the performance of local models against global models in noisy conditions.

Main Methods:

• Dividing the search space into non-overlapping hyperspheres (local models).
• Utilizing a direction of improvement based on average sphere performance to guide solutions.
• Implementing and testing the local model framework on a diverse set of noisy problems.

Main Results:

• The local models framework demonstrated superior noise filtering capabilities.
• The proposed approach significantly increased the robustness of the evolutionary algorithm in noisy settings.
• Experimental results showed better performance compared to traditional global models.

Conclusions:

• Local models offer an effective strategy for mitigating noise in EMO.
• The framework enhances the reliability and performance of evolutionary algorithms in real-world noisy applications.
• This approach provides a robust alternative for tackling noisy multi-objective optimization problems.