A Surrogate-Assisted Multiswarm Optimization Algorithm for High-Dimensional Computationally Expensive Problems | JoVE Visualize

Area of Science:

Computational Intelligence
Optimization Algorithms
Machine Learning

Background:

High-dimensional and computationally expensive problems pose significant challenges in various scientific and engineering fields.
Existing optimization algorithms often struggle with balancing exploration and exploitation in complex search spaces.
The need for efficient algorithms that can handle computationally intensive tasks is critical.

Purpose of the Study:

To introduce a novel Surrogate-Assisted Multi-Swarm Optimization (SAMSO) algorithm.
To address the limitations of current methods in solving high-dimensional, computationally expensive optimization problems.
To enhance the efficiency and effectiveness of optimization through synergistic swarm interactions.

Main Methods:

The SAMSO algorithm integrates two swarms: Teaching-Learning-Based Optimization (TLBO) for enhanced exploration and Particle Swarm Optimization (PSO) for rapid convergence.
A dynamic swarm size adjustment scheme is implemented to manage evolutionary progress.
A novel prescreening criterion is introduced to optimize the selection of individuals for exact function evaluations.
Dual coordinate systems are employed to improve PSO's search efficiency across diverse function landscapes.

Main Results:

The proposed SAMSO algorithm demonstrated superior performance compared to three state-of-the-art algorithms.
Experiments conducted on benchmark functions with dimensions ranging from 30 to 200 validated the algorithm's effectiveness.
The synergistic learning between the TLBO and PSO swarms significantly improved search capabilities.
The dynamic swarm size adjustment and prescreening criterion contributed to efficient evolutionary progress.

Conclusions:

The SAMSO algorithm offers a powerful and efficient solution for high-dimensional, computationally expensive optimization problems.
The integration of TLBO and PSO, coupled with adaptive mechanisms, provides a robust framework for complex optimization tasks.
SAMSO represents a significant advancement in surrogate-assisted optimization, outperforming existing methods in benchmark evaluations.