An Attention Learning Particle Swarm Optimizer with Sensitive Search Transition

Particle swarm optimization (PSO) has been extensively applied to diverse optimization problems, yet it may encounter challenges such as premature convergence and slow convergence, especially in complex optimization problems. Thus, how to guide particles away from local optima and exploit promising areas promptly becomes important for improving the performance of PSO. Considering the above issues, we propose an attention learning particle swarm optimizer with sensitive search transition (ALPSO-SST). In ALPSO-SST, the specially designed attentional learning (AL) strategy allocates higher attention to learning targets with superior quality, so that the learning targets are integrated into an attentional particle with these diverse attentions. Such an attentional particle guides the population toward superior areas. Additionally, through employing the principle of probability distribution entropy, the evolution level of the population can be sensitively captured by a specially designed evolution indicator, enabling a sensitive search transition (SST) from exploration to exploitation. Experimental results demonstrate that ALPSO-SST outperforms five PSO-based algorithms across eight well-known test functions in both convergence speed and solution quality, offering promising prospects for future research and applications in real-world optimization problems.