<p>This paper proposes the multiplier leadership optimization algorithm, which draws inspiration from multiplier leadership principles to search for and optimize solutions to complex problems effectively. Multiplier leaders possess the unique ability to amplify their teams’ collective intelligence and capabilities. They cultivate an environment of open discussion, creative thinking, accountability, and motivating team members to excel. The proposed algorithm is implemented in MATLAB, and its performance is evaluated on the IEEE Congress on Evolutionary Computation 2021 test bench suite, which consists of 80 benchmark functions. The performance of the proposed approach is compared with those of the other 9 other state-of-the-art optimization algorithms. The proposed algorithm successfully solved 47 out of 80 benchmark functions, demonstrating its superior performance. The computational complexity of proposed algorithm is <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="43926_2025_168_Article_IEq1.gif" Format="GIF" Height="20" Rendition="HTML" Resolution="72" Type="Linedraw" Width="83" /> </InlineMediaObject> <EquationSource Format="TEX">$O(M * N^2)$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>O</mi> <mo stretchy="false">(</mo> <mi>M</mi> <mrow /> <mo>∗</mo> <msup> <mi>N</mi> <mn>2</mn> </msup> <mo stretchy="false">)</mo> </mrow> </math></EquationSource> </InlineEquation>, making it efficient for large-scale problems. Furthermore, this paper proposes an MLOA-based approach to evolve the near-optimal architecture of convolution neural networks for Melanoma classification. The proposed approach is implemented in Python and tested on the PH2 and ISIC2020 datasets. The PH2 dataset consists of 200 dermoscopic images, while the ISIC2020 dataset comprises 33,126 images. The proposed approach is compared with 12 other state-of-the-art methods and achieves high classification accuracies of 98.97% on the PH2 dataset and 99.47% on the ISIC2020 dataset. Additional performance metrics include precision of 98.72%, recall of 98.85%, and F1-score of 98.78% on the PH2 dataset, while on the ISIC2020 dataset, it achieves precision of 99.32%, recall of 99.51%, and F1-score of 99.41%. The results indicate that MLOA-based CNN architecture outperforms all existing approaches for Melanoma classification, offering a promising solution for medical image analysis.</p>

Multiplier leadership optimization algorithm (MLOA): unconstrained global optimization approach for melanoma classification

期刊论文
出版模式：

开放获取
发表日期：

2025年6月7日

查看全文

Sukanta Ghosh,
Amar Singh,
Shakti Kumar

摘要

This paper proposes the multiplier leadership optimization algorithm, which draws inspiration from multiplier leadership principles to search for and optimize solutions to complex problems effectively. Multiplier leaders possess the unique ability to amplify their teams’ collective intelligence and capabilities. They cultivate an environment of open discussion, creative thinking, accountability, and motivating team members to excel. The proposed algorithm is implemented in MATLAB, and its performance is evaluated on the IEEE Congress on Evolutionary Computation 2021 test bench suite, which consists of 80 benchmark functions. The performance of the proposed approach is compared with those of the other 9 other state-of-the-art optimization algorithms. The proposed algorithm successfully solved 47 out of 80 benchmark functions, demonstrating its superior performance. The computational complexity of proposed algorithm is $O(M * N^2)$ $O (M * N^{2})$ , making it efficient for large-scale problems. Furthermore, this paper proposes an MLOA-based approach to evolve the near-optimal architecture of convolution neural networks for Melanoma classification. The proposed approach is implemented in Python and tested on the PH2 and ISIC2020 datasets. The PH2 dataset consists of 200 dermoscopic images, while the ISIC2020 dataset comprises 33,126 images. The proposed approach is compared with 12 other state-of-the-art methods and achieves high classification accuracies of 98.97% on the PH2 dataset and 99.47% on the ISIC2020 dataset. Additional performance metrics include precision of 98.72%, recall of 98.85%, and F1-score of 98.78% on the PH2 dataset, while on the ISIC2020 dataset, it achieves precision of 99.32%, recall of 99.51%, and F1-score of 99.41%. The results indicate that MLOA-based CNN architecture outperforms all existing approaches for Melanoma classification, offering a promising solution for medical image analysis.