Anchovy-inspired filter algorithm: A bio-inspired optimization approach for high-dimensional benchmark functions

Abstract

This paper presents the anchovy-inspired filter algorithm (AFA), a novel bio-inspired metaheuristic optimization method motivated by the filter feeding behavior of anchovies. Unlike conventional swarm intelligence algorithms, AFA employs a filtering mechanism in which each agent generates multiple candidate solutions within a local sampling radius and selects the best, mimicking how anchovies filter microscopic prey from seawater. To evaluate its performance, AFA was benchmarked against particle swarm optimization (PSO) and genetic algorithm (GA) using six standard test functions: Sphere, Rosenbrock, Schwefel 1.2, Rastrigin, Griewank, and Ackley in 30-dimensional search spaces. Simulation results demonstrate that AFA consistently outperforms PSO and GA across unimodal and multimodal functions. For unimodal problems such as Sphere, Rosenbrock, and Schwefel 1.2, AFA achieved significantly lower best and mean fitness values, reflecting strong exploitation capability. For multimodal functions including Rastrigin, Griewank, and Ackley, AFA effectively avoided local minima, maintained robustness, and achieved stable convergence with lower variance. Convergence analysis further indicates that AFA steadily approaches near-global optima without premature stagnation. Overall, the results highlight the effectiveness of the filter-based exploitation mechanism in balancing exploration and exploitation. Future research will focus on adaptive filtering strategies, hybrid integration with other metaheuristics, and applications to real-world optimization problems.