NPROS: A Not So Pure Random Orthogonal search algorithm—A suite of random optimization algorithms driven by reinforcement learning

• Published in: Optimization letters Vol. 18; no. 9; pp. 2091 - 2111
• Main Authors: Hameed, A. S. Syed Shahul, Rajagopalan, Narendran
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2024
• Subjects: Computational Intelligence; Mathematics; Mathematics and Statistics; Numerical and Computational Physics; Operations Research/Decision Theory; Optimization; Original Paper; Simulation; Global optimization; Metaheuristics; Multi-armed bandit; Reinforcement learning; Random search
• ISSN: 1862-4472; 1862-4480
• DOI: 10.1007/s11590-023-02038-0

We live in a world where waves of novel nature-inspired metaheuristic algorithms keep hitting the shore repeatedly. This never-ending surge of new metaheuristic algorithms is overwhelming to the extent that their novelty is being criticized. In this paper, instead of focusing on metaheuristics, we focus on pure random search algorithms for global optimization. Pure Random Orthogonal Search (PROS) is a recently published random optimization algorithm, which is strikingly simple, involves no parameter tuning, but is very effective in solving global optimization problems. In this paper, we propose a modified version of the PROS algorithm, which injects a flavor of exploitation into the otherwise purely explorative PROS algorithm. Further, the concepts of reinforcement learning are utilized to provide the proposed algorithm the ability to ‘learn’ to take the optimal actions, to find the global optima. The source code of NPROS is publicly available at: https://github.com/Shahul-Rahman/NPROS