analiza aiqbench
This is an incremental metaheuristic paper that adds four modifications to an existing wave-optics-inspired optimizer, aimed at slow convergence and premature stagnation. Its evidence is benchmark Friedman ranks on standard CEC suites against a restricted set of comparison algorithms, plus six constrained engineering problems, so the gains are relative and specific to those test sets. For AI readers, it is a useful data point on evolutionary and nature-inspired search methods, but it does not show that the approach transfers to modern ML training or large-scale problems.
IWOO: Wave Optics Optimizer Improved with Hybrid Strategies for Numerical Optimization and Engineering Applications
arXiv:2610.23994v1 Announce Type: cross Abstract: The Wave Optics Optimizer (WOO) incorporates optical diffraction and interference into numerical optimization and has shown competitive performance in benchmarks and engineering applications. However, it can converge slowly, become trapped in local optima, and lose search effectiveness in later iterations. This paper proposes an Improved Wave Optics Optimizer (IWOO) that integrates four complementary strategies. First, a good point set and opposition-based learning are combined in GPS--OBL initialization to improve population quality and diversity. Second, an E-DM nonlinear perturbation mechanism is embedded in temporary triangle-based search to develop the TriEDM multi-path strategy and enhance search capability. Third, Correlation-Guided Groupwise Orthogonal Learning (CGOL) leverages relationships among decision variables to improve search efficiency while reducing computational overhead. Fourth, an adaptive stagnation-aware Runge--Kutta correction is introduced to mitigate premature convergence. IWOO is evaluated on the IEEE CEC2019, CEC2022, and CEC2017 benchmark suites against representative and high-performing optimizers, with ablation studies assessing the contribution of each strategy. Within the representative algorithm group, IWOO achieves Friedman ranks of 1.8000, 1.5833, and 1.4483 on CEC2019, CEC2022, and CEC2017, respectively. Tests on six constrained engineering optimization problems further demonstrate its competitive accuracy, stability, and practical applicability.
Abstract:The Wave Optics Optimizer (WOO) incorporates optical diffraction and interference into numerical optimization and has shown competitive performance in benchmarks and engineering applications. However, it can converge slowly, become trapped in local optima, and lose search effectiveness in later iterations. This paper proposes an Improved Wave Optics Optimizer (IWOO) that integrates four complementary strategies. First, a good point set and opposition-based learning are combined in GPS--OBL initialization to improve population quality and diversity. Second, an E-DM nonlinear perturbation mechanism is embedded in temporary triangle-based search to develop the TriEDM multi-path strategy and enhance search capability. Third, Correlation-Guided Groupwise Orthogonal Learning (CGOL) leverages relationships among decision variables to improve search efficiency while reducing computational overhead. Fourth, an adaptive stagnation-aware Runge--Kutta correction is introduced to mitigate premature convergence. IWOO is evaluated on the IEEE CEC2019, CEC2022, and CEC2017 benchmark suites against representative and high-performing optimizers, with ablation studies assessing the contribution of each strategy. Within the representative algorithm group, IWOO achieves Friedman ranks of 1.8000, 1.5833, and 1.4483 on CEC2019, CEC2022, and CEC2017, respectively. Tests on six constrained engineering optimization problems further demonstrate its competitive accuracy, stability, and practical applicability.
Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)
arXiv:2610.23994 [cs.NE]
(or arXiv:2610.23994v1 [cs.NE] for this version)
https://doi.org/10.48550/arXiv.2610.23994 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 285 cuvinte
Subiecte conexe
Reinforcement Learning-Guided NSGA-II Enhanced with Gray Relational Coefficient for Multi-Objective Optimization: Application to NASDAQ Portfolio Optimization
arXiv · 21 iul. 2026
Performance Profiling on AMD GPUs - Part 4: Fortran OpenMP Offload Edition
AMD · 1 iun. 2026
Scalable Variational Quantum Optimization via Pauli Correlation Encoding: Application to Large-Scale Power Demand Portfolio Optimization
arXiv · 28 iul. 2026
Distributed Quantum Approximate Optimization Algorithm on a Quantum-Centric Supercomputing Architecture
arXiv · 23 iul. 2026