A Hybrid Multi-Strategy Differential Creative Search Optimization Algorithm and Its Applications.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Traditional divergent quantum-inspired differential search (DCS) algorithms suffer from uneven population distribution and limited search accuracy.
• Addressing these limitations is crucial for enhancing the performance of optimization algorithms.

Purpose of the Study:

• To propose a hybrid multi-strategy variant, DQDCS, to overcome the limitations of the traditional DCS algorithm.
• To improve population diversity, escape local optima, and balance exploration-exploitation for faster convergence.

Main Methods:

• Implemented a refined set strategy and clustering process for population initialization, replacing pseudo-random methods.
• Introduced a novel position update mechanism to facilitate escaping local optima.
• Integrated a double Q-learning model to balance exploration and exploitation probabilities.

Main Results:

• Ablation studies validated the effectiveness of each enhancement.
• The Wilcoxon rank-sum test confirmed the statistical significance of DQDCS's performance.
• Benchmark simulations on CEC2019, CEC2022 test functions, and engineering design problems demonstrated superior performance.

Conclusions:

• The DQDCS algorithm significantly improves convergence speed and optimization precision compared to classical and state-of-the-art algorithms.
• The integration of refined set, clustering, and double Q-learning mechanisms is key to the enhanced performance.