All-dimension neighborhood based particle swarm optimization with randomly selected neighbors

Particle swarm optimization (PSO) is widely used for solving various optimization problems, since it has few parameters and is easy to implement. However, canonical PSO generally suffers from premature convergence because it usually loses diversity too rapidly during the evolutionary process. To improve the performance of PSO on complex problems, an all-dimension-neighborhood-based PSO with randomly selected neighbors learning strategy (ADN-RSN-PSO) is proposed in this study. The randomly selected neighbors (RSN) learning strategy is adopted in the early stage of PSO to enhance the swarm diversity, while the all-dimension neighborhood (ADN) strategy is utilized in the later stage to accelerate the convergence rate. The ADN strategy enhances the local search capability around the global-best solution in a dimension by dimension manner, and the search distance is adapted by shrinking and random-expansion operators. Experimental results show that ADN-PSO can improve the exploitation capability of the global version of PSO. To test the performance of the proposed ADN-RSN-PSO, comparison tests on the CEC2013 test suite are carried out. The comparison results reveal that ADN-RSN-PSO outperforms other peer PSO variants. In the end, the proposed ADN-RSN-PSO is applied to the radar system design problem to demonstrate its potential in real-life applications.