Stochastic Particle Swarm Optimization and its variants for Multimodal Function Optimization

The particle swarm optimization (PSO) is a stochastic, population-based optimization algorithm. The PSO can be applied to the wide range of engineeringfields. This work presents an improved particle swarm optimization using the stochastic local search concept (SPSO), employing dynamic inertia weight tosignificantly improve the performance of basic PSO algorithm. Under this method, to balance between exploration and exploitation, at each iteration step, a blob is associated with each candidate particle, and a local exploration performed in this blob. The stochasticlocal search encourages the particle to explore this blob beyond that defined by the search algorithm to achieve better solution. Over the successive iterations, the blobsize dynamically decreases. To further improve performance of the proposed approach a non-linear dynamic inertia weight introduced. SPSO variations tested on a commonly used set of multimodal functions. Experimental results show that SPSO is effective androbust, and outperforms other algorithms investigated in this consideration

Keywords:

Optimization , Particle swarm optimization , Stochastic local search , Multimodal functions