k-NNC: a Simple Classification Model for Reducing Computational Volumn in Metaheuristic Optimization

EasyChair Preprint 12642

Abstract

In recent years, metaheuristic (MH) optimization algorithms have been increasingly applied in engineering optimal designs due to their ability to search for global solution and solve problems with a large number of variables. design. However, the disadvantage of MH is the large amount of computation because MH often requires thousands of evaluations of objective function and constraints. Recently, the k-nearest neighbor comparison (k-NNC) method has been proposed to reduce computational costs when performing optimization using MH. k-NNC considers a new design solution by comparing its k closest existing designs (k-nearest neighbors) with another design in the population. The new design will be rejected without performing an evaluation if the majority of the k nearest neighboring designs are inferior to the compared design. k-NNC has been combined with Rao algorithms to optimize the weight of truss structures. As shown through numerical examples, k-NNC significantly reduces the number of structural analyses. In this paper, the capabilities of k-NNC are confirmed when combined with some other popular MH algorithms such as differential evolution and Jaya. The results when applied to some engineering optimization problems have proven that k-NNC is a simple and effective model to save computational costs for MH.

Keyphrases: Metaheuristic, k-NNC, mô hình phân loại, thiết kế tối ưu

@booklet{EasyChair:12642,
  author    = {Hoang-Anh Pham and Manh-Hung Ha},
  title     = {k-NNC: a Simple Classification Model for Reducing Computational Volumn in Metaheuristic Optimization},
  howpublished = {EasyChair Preprint 12642},
  year      = {EasyChair, 2024}}