Effects of widely separated clusters on lotto-type competitive learning with particle swarm features

This correspondence describes our attempts of incorporating particle swarm features into competitive learning. We first outline our reinterpretation of the symbols and notations used in particle swarm optimisation (PSO) algorithms. Three versions of modifications to the classical frequency-sensitive competitive learning are presented. A new contraction/expansion phenomenon is illustrated. We then examine the effect of introducing particle swarm like features in our lotto-type competitive learning. Experimental results indicate that, like the PSO algorithms, a careful selection of the values for the control parameters is necessary for the successful convergence of particles. With the new modifications, we show experimentally that the modified algorithm can behave both similar to PSO algorithms and the original lotto-type competitive learning algorithms.