Considering radial basis function neural network for effective solution generation in metaheuristic algorithms

Area of Science:

• Engineering Optimization
• Computational Intelligence
• Machine Learning

Background:

• Engineering optimization problems often face severe limitations on function evaluations due to time and cost constraints.
• Existing metaheuristic methods typically require numerous function evaluations, posing a challenge for global optimization.
• Efficiently finding optimal solutions under evaluation constraints is a critical research area.

Purpose of the Study:

• To present a new metaheuristic optimization algorithm designed to significantly reduce function evaluations.
• To leverage radial basis function neural networks (RBFNN) for guiding the optimization search process.
• To enhance the efficiency and effectiveness of global optimization in computationally constrained environments.

Main Methods:

• The proposed algorithm strategically distributes initial solutions using a maximum design approach.
• A radial basis function neural network (RBFNN) models objective function values from current solutions.
• Key neurons in the RBFNN's hidden layer identify promising search regions, guiding new solution generation via centroids and standard deviations.

Main Results:

• The algorithm effectively reduces the number of function evaluations by focusing on high-value objective function areas.
• Comparative analysis across test functions shows consistent outperformance against popular metaheuristic algorithms.
• The new method demonstrates improved convergence rates and delivers higher-quality solutions.

Conclusions:

• The developed metaheuristic optimization algorithm offers a significant reduction in function evaluations.
• The integration of RBFNN provides an effective mechanism for guiding the search process in constrained optimization.
• This approach presents a promising advancement for tackling complex engineering optimization challenges with limited resources.