Adaptive Clubs-based Particle Swarm Optimization

This paper introduces a new dynamic neighborhood network for particle swarm optimization. In Club-based Particle Swarm Optimization (C-PSO) algorithm, each particle initially joins a default number of social groups (clubs). Each particle is affected by its own experience and the experience of the best performing member of the social groups it is a member of. In the proposed Adaptive membership C-PSO (AMC-PSO), a time varying default Membership is introduced. This modification enables the particles to explore the space based on their own experience in the first stage, and to intensify the connections of the social network in later stages to avoid premature convergence. This proposed dynamic neighborhood algorithm is compared with other PSO algorithms having both static and dynamic neighborhood topologies on a set of classic benchmark problems. The results showed superior performance for AMC-PSO regarding its ability to escape from local optima, while its speed of convergence is comparable to other algorithms.