Red Snapper Optimization Algorithm

To address the problems of premature convergence, exploration-exploitation imbalance, and insufficient utilization of local structures in traditional swarm intelligence algorithms for high-dimensional, multi-peaked, and strongly nonlinear continuous optimization problems, a red seabream optimization algorithm based on information refraction field and entropy-driven phase transition mechanism is proposed. This algorithm abstracts five dynamic mechanisms from red seabream's deep-water migration, reef attachment, group perception, and sudden predation behavior: information refraction field migration mechanism, deep-risk dual-potential energy system, reef memory manifold compression mechanism, group entropy-driven phase transition mechanism, and risk refraction sprint mechanism. By constructing a refraction vector modulated by curvature signs, the nonlinear search trajectory is curved; by establishing a time-decay weighted dual-potential energy model, a unified energy framework for exploration and exploitation is achieved; by constructing a low-dimensional submanifold local search through principal direction projection, the efficiency of high-dimensional problems is improved; by using fitness distribution entropy, stage adaptive switching is achieved; and by triggering a refraction-type long-jump escape from local extrema through a risk accumulation criterion. Theoretical analysis shows that the algorithm satisfies the irreducibility of weak Markov chains and the non-zero neighborhood arrival probability under random perturbation and bounded energy coefficient conditions, thus possessing probabilistic convergence properties. This method provides a new approach for constructing a swarm intelligence optimization model with a unified energy interpretation.