Performance Analysis of the Artificial Physics Optimization Algorithm with Simple Neighborhood Topologies

Abstract

As a novel population-based optimization technique, the artificial physics optimization (APO) algorithm inspired by physics is presented recently. Although it is characteristic of rapid convergence speed, it also suffers from worse diversity and premature convergence. Accordingly, drawing lessons from those strategies for interactions among individuals in other algorithms, this paper presents the local artificial physics optimization (LAPO) algorithm both to apply it under some simply topologies and to get an insight into the effect of structures. For this end, the performances of LAPO algorithm under particular topologies are investigated by the gravitation constant G adjusting. Simulation results show that LAPO algorithm is valid under some neighborhood structures and that the gravitation constant G has a great influence on performance of LAPO algorithm with different topologies. Also, the results from simulation indicate that the presented LAPO algorithm is superior to APO algorithm so long as parameter G is selected properly under particular topologies.