Scholarlink

<p>Random Optimization is an under-explored subdomain of optimization. Nature-inspired metaheuristics have always managed to solicit the majority of attention, and as a repercussion, interest in other kinds of optimization algorithms got deprived. Post the initial buzz during the 60 &amp; the 70s, the field of random optimization became quite dormant. However, the recent introduction of the Pure Random Orthogonal Search (PROS) algorithm has managed to revive the lost interest in random optimization. PROS algorithm is strikingly simple, hyperparameter free, and easier to implement, but can deliver optimization performance in lieu of established classical optimization algorithms. A discussion on the taxonomy of different random optimization methods and the mathematical properties of PROS is presented to better understand its functionality. We propose a modification to the PROS algorithm, which we name iPROS (Improved PROS). Random optimization is mainly preferred for its elementary and uncomplicated nature. Adhering to this philosophy, the proposed modification is a minor intuitive change that does not distort the elegance of the original algorithm. The modification is to progressively shrink the feasible region within which the search for the global optimum occurs. Extensive experimentation with classical benchmark functions, shows that the proposed modification is superior to the original algorithm. The source code of iPROS is publicly available at <a href="https://github.com/Shahul-Rahman/iPROS">https://github.com/Shahul-Rahman/iPROS</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

iPROS: improved pure random orthogonal search for global optimization with active shrinkage of feasible region

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • A. S. Syed Shahul Hameed,
  • R. Allwin,
  • Manindra Narayan Singh,
  • Animesh Nanda,
  • Narendran Rajagopalan

摘要

Random Optimization is an under-explored subdomain of optimization. Nature-inspired metaheuristics have always managed to solicit the majority of attention, and as a repercussion, interest in other kinds of optimization algorithms got deprived. Post the initial buzz during the 60 & the 70s, the field of random optimization became quite dormant. However, the recent introduction of the Pure Random Orthogonal Search (PROS) algorithm has managed to revive the lost interest in random optimization. PROS algorithm is strikingly simple, hyperparameter free, and easier to implement, but can deliver optimization performance in lieu of established classical optimization algorithms. A discussion on the taxonomy of different random optimization methods and the mathematical properties of PROS is presented to better understand its functionality. We propose a modification to the PROS algorithm, which we name iPROS (Improved PROS). Random optimization is mainly preferred for its elementary and uncomplicated nature. Adhering to this philosophy, the proposed modification is a minor intuitive change that does not distort the elegance of the original algorithm. The modification is to progressively shrink the feasible region within which the search for the global optimum occurs. Extensive experimentation with classical benchmark functions, shows that the proposed modification is superior to the original algorithm. The source code of iPROS is publicly available at https://github.com/Shahul-Rahman/iPROS.