A neutrosophic stagnation-gated framework for population-based optimizers

Metaheuristic algorithms often struggle to balance exploration and exploitation. They may become trapped in local optima. This paper proposes a novel, generic framework that combines neutrosophic set theory with a stagnation-gated local escape mechanism. The goal is to improve convergence performance for many swarm-based and evolutionary optimizers. At each iteration, every agent’s fitness is mapped to three independent degrees: truth (T), falsity (F), and indeterminacy (I). These are normalized over the population. The degrees dynamically adapt key algorithm parameters, such as inertia and acceleration coefficients in PSO. A stagnation counter tracks the number of consecutive non-improving iterations per agent. Only agents with both high indeterminacy (I > Ithresh) and prolonged stagnation (count > Slimit) trigger a controlled local escape. This step perturbs their position with a small, fitness-weighted random offset. Tests on standard benchmarks show that our approach accelerates convergence, improves solutions, and robustly avoids local traps. This is achieved with minimal parameter tuning. The framework provides a modular enhancement that joins single-valued neutrosophic membership degrees with a stagnation-gated local escape mechanism. It regulates exploration and exploitation at the agent level. Extensive experiments on standard benchmarks show statistically significant gains in convergence and solution quality for several population-based optimizers, especially PSO, GWO, and WOA, compared to their baselines and similar methods. Nonparametric statistics confirm the robustness of these gains. However, results for DE and SCA highlight some limitations that may require algorithm-specific integration strategies. The modular framework can be added to existing metaheuristics to yield consistent performance improvements. Our implementation is publicly available at https://github.com/kayzersoze۲۶/ns_mthrs, enabling reproducibility and facilitating further research.

کلیدواژه ها:

Metaheuristic Algorithm ، neutrosophic set ، local escape mechanism ، population-based algorithm