A Distributed Swarm Optimizer With Adaptive Communication for Large-Scale Optimization.

Area of Science:

• Computational Science
• Optimization Algorithms
• Distributed Computing

Background:

• Large-scale optimization problems are increasingly common, presenting significant computational challenges.
• Existing methods struggle with high dimensionality and computational cost.
• Distributed evolutionary computation is crucial for efficient problem-solving.

Purpose of the Study:

• To propose a novel distributed swarm optimizer for tackling large-scale, high-dimensional optimization problems.
• To enhance computational efficiency and solution quality compared to existing methods.
• To enable parallel processing for complex optimization tasks.

Main Methods:

• A master-slave distributed model where a master manages communication and slaves iterate swarms.
• An asynchronous and adaptive communication strategy based on a request-response mechanism.
• An elite-guided learning strategy using current and historical best solutions to update particles.

Main Results:

• The distributed optimizer achieves competitive solution quality against state-of-the-art methods.
• Significant acceleration in execution time compared to sequential algorithms, with near-linear speedup.
• Demonstrated good scalability for solving higher-dimensional problems.

Conclusions:

• The proposed distributed swarm optimizer effectively addresses challenges in large-scale optimization.
• The method offers significant computational speedups and maintains high solution quality.
• The algorithm exhibits excellent scalability for high-dimensional and computationally intensive problems.