Fractional particle swarm optimization in multidimensional search space.

Area of Science:

• Computational Intelligence
• Optimization Algorithms

Background:

• Particle Swarm Optimization (PSO) faces challenges with high-dimensional, multimodal problems.
• Existing PSO variants struggle with premature convergence and lack of divergence in complex search spaces.

Purpose of the Study:

• To introduce two novel techniques, Multidimensional (MD) PSO and Fractional Global Best Formation (FGBF), to address limitations in Particle Swarm Optimization.
• To enhance the ability of swarm optimizers to find optima in high-dimensional, complex multimodal search spaces without prior dimension specification.

Main Methods:

• Multidimensional (MD) PSO: Re-structures swarm particles for interdimensional passes, enabling simultaneous search for positional and dimensional optima.
• Fractional Global Best Formation (FGBF): Creates an artificial global best (aGB) particle using fractional components of best dimensional elements to improve guidance and diversity.

Main Results:

• MD PSO with FGBF demonstrates significant speed gains in nonlinear function minimization and data clustering.
• The combined approach converges to global optima at the true dimension, irrespective of search space dimension, swarm size, or problem complexity.
• Experiments show improved performance over existing PSO variants on complex multimodal problems.

Conclusions:

• MD PSO with FGBF offers a robust and efficient solution for high-dimensional, complex multimodal optimization problems.
• The proposed techniques overcome the limitations of fixed dimension a priori and premature convergence inherent in traditional PSO.
• This breakthrough enhances the applicability of swarm intelligence in challenging optimization domains.