Triple Archives Particle Swarm Optimization.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) faces challenges in exemplar selection and designing efficient learning models.
• Existing PSO variants often struggle with balancing exploration and exploitation effectively.

Purpose of the Study:

• To introduce a novel Triple Archives Particle Swarm Optimization (TAPSO) algorithm.
• To address the limitations of exemplar selection and learning model design in PSO.
• To enhance solution accuracy and convergence speed in optimization tasks.

Main Methods:

• TAPSO utilizes three archives: one for elites (best fitness), one for profiteers (fast progress), and one for outstanding exemplars.
• Potential exemplars are bred using a pair of elite and profiteer parents.
• Particles employ learning models based on their potential exemplars' fitness, simplifying models by omitting acceleration coefficients.
• Outstanding exemplars are saved for reuse by inferior particles to boost exploitation and conserve resources.

Main Results:

• TAPSO demonstrated very promising performance across 30 benchmark functions and four real-world applications.
• Comparisons with eight other PSO variants showed TAPSO achieving superior solution accuracy.
• TAPSO consistently exhibited faster convergence speeds compared to existing methods.

Conclusions:

• TAPSO effectively addresses key challenges in PSO, namely exemplar selection and learning model design.
• The proposed strategies enhance both the exploitation capability and computational efficiency of the algorithm.
• TAPSO offers a robust and efficient alternative for various optimization problems.