Enhanced Comprehensive Learning Particle Swarm Optimization with Dimensional Independent and Adaptive Parameters

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Metaheuristics

Background:

• Particle Swarm Optimization (PSO) is a metaheuristic for global optimization.
• Comprehensive Learning PSO (CLPSO) and Enhanced CLPSO (ECLPSO) are advanced variants.
• ECLPSO improves exploitation and convergence but struggles with exploration for certain problems.

Purpose of the Study:

• To enhance the exploration capabilities of ECLPSO.
• To develop a novel metaheuristic that addresses ECLPSO's limitations in finding global optima.
• To improve the performance of PSO variants in complex optimization tasks.

Main Methods:

• Proposing Adaptive CLPSO (ACLPSO), a novel metaheuristic.
• Assigning independent inertia weights, acceleration coefficients, and learning probabilities per dimension and particle.
• Adaptively updating dimensional parameters based on normative intervals to balance exploration and exploitation.

Main Results:

• ACLPSO demonstrates significantly improved exploration performance compared to ECLPSO.
• The proposed dimensional independent and adaptive parameters effectively enhance optimization.
• ACLPSO successfully derives global or near-optimum solutions across various benchmark functions.

Conclusions:

• ACLPSO effectively overcomes the exploration limitations of previous CLPSO variants.
• The adaptive, dimension-specific parameter tuning is crucial for superior performance.
• ACLPSO offers a robust and effective approach for global optimization problems.