A deep memory bare-bones particle swarm optimization algorithm for single-objective optimization problems.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Swarm intelligence

Background:

• Single-objective optimization problems present challenges in balancing global search and local accuracy.
• Traditional particle swarm optimizers often suffer from premature convergence and entrapment in local optima due to loss of diversity.

Purpose of the Study:

• To introduce a novel deep memory bare-bones particle swarm optimization algorithm (DMBBPSO).
• To enhance the performance of particle swarm optimization for high-dimensional, complex single-objective problems.

Main Methods:

• The DMBBPSO integrates a multiple memory storage mechanism (MMSM) to increase swarm diversity.
• A layer-by-layer activation strategy (LAS) is employed to prevent premature convergence and improve local search.
• Both personal best positions and deep memories inform particle evaluation.

Main Results:

• Experiments using CEC2017 benchmark functions demonstrate the DMBBPSO's effectiveness.
• The DMBBPSO achieved high-precision results, outperforming five state-of-the-art evolutionary algorithms.
• The algorithm showed enhanced robustness and avoided premature convergence.

Conclusions:

• The DMBBPSO effectively addresses the challenge of maintaining diversity and accuracy in global searches.
• The proposed algorithm offers a reliable and robust solution for complex single-objective optimization problems.
• The combination of MMSM and LAS significantly improves optimization capabilities.