A Survey on GPU-Based Implementation of Swarm Intelligence Algorithms.

Area of Science:

• Computer Science
• Artificial Intelligence
• Optimization

Background:

• Swarm intelligence algorithms (SIAs) are inspired by natural collective behavior for optimization.
• Complex problems require numerous fitness function evaluations, hindering SIA efficiency.
• Graphical Processing Units (GPUs) offer parallel processing capabilities to accelerate SIAs.

Purpose of the Study:

• To comprehensively review GPU-based parallel SIAs.
• To propose a novel taxonomy for categorizing these algorithms.
• To introduce new criteria for evaluating parallel implementation and performance.

Main Methods:

• A systematic review of existing literature on GPU-based SIAs.
• Development of a new taxonomy for GPU-accelerated SIAs.
• Proposal of universal criteria for performance evaluation.
• Case studies to validate the proposed methodology and criteria.

Main Results:

• A comprehensive review and taxonomy of GPU-based parallel SIAs are presented.
• Critical factors for efficient parallel implementation are detailed.
• Novel criteria for performance evaluation are proposed and validated.
• The effectiveness of the proposed methodology is demonstrated through case studies.

Conclusions:

• GPU acceleration significantly enhances SIA performance for complex optimization problems.
• The proposed taxonomy and evaluation criteria provide a standardized framework for research.
• Future research directions in GPU-based SIAs are outlined.