A Neighborhood Regression Optimization Algorithm for Computationally Expensive Optimization Problems.

Area of Science:

• Computational Mathematics
• Optimization Theory
• Algorithm Development

Background:

• Computationally expensive optimization problems present significant challenges for global optimization due to high function evaluation costs.
• Existing algorithms often struggle with efficiency and performance when dealing with such problems across diverse fields.

Purpose of the Study:

• To propose a novel, simple, and effective optimization algorithm for computationally expensive problems.
• To address the challenge of high computational cost in global optimization.

Main Methods:

• Introduced the neighborhood regression optimization algorithm.
• Employs a regression technique within a neighborhood structure to predict descent directions for minimization problems.
• Generates new potential solutions (offspring) around the current best solution using the predicted descent direction.

Main Results:

• The proposed algorithm demonstrated clear advantages on unimodal and smooth problems.
• It showed competitive or superior overall performance compared to 12 popular algorithms across benchmark suites with up to 30 decision variables.
• The algorithm proved efficient, balancing solution quality with running time.

Conclusions:

• The neighborhood regression optimization algorithm is a viable and effective approach for computationally expensive optimization tasks.
• Its performance indicates strong potential for practical applications where function evaluations are costly.
• The algorithm offers a good balance between solution accuracy and computational efficiency.