Particle swarm optimization with deliberate loss of information

We introduce a new variant for the constriction coefficient model of the established particle swarm optimization (PSO) algorithm. The new variant stands between the synchronous and asynchronous version of PSO, combining their operation regarding the update and evaluation frequency of the particles. Yet, the proposed variant has a unique feature that distinguishes it from other approaches. Specifically, it allows the undisrupted move of all particles even though evaluating only a portion of them. Apparently, this implies a loss of information for PSO, but it also allows the full exploitation of the convergence dynamic of the constriction coefficient model. Moreover, it requires only minor modifications to the original PSO algorithm since it does not introduce complicated procedures. Experimental results on widely used benchmark problems as well as on problems drawn from real-life applications, reveal that the proposed approach is efficient and can be very competitive to other PSO variants as well as to more specialized approaches.