Impact of Acceleration Coefficients and Random Neighborhood Topology Modification Threshold on SPSO

Particle Swarm Optimization is motivated by collective behavior in the nature. Earlier PSO variants were biased and rotationally variant because of dimension by dimension method. Clerc proposed Standard PSO 2011 version to overcome the bias which modifies velocity in a geometrical way. Particle Swarm Optimization is an optimization technique having few parameters to tune. The neighborhood topology and acceleration coefficients are two major parameters to enhance PSO performance. The work under consideration extends the existing parameter setting techniques on acceleration coefficients. The concept of a "dynamic" neighborhood is explored here. The paper focuses on random and linear increment and decrement acceleration coefficients as well as random neighborhood topology link based variants. Experiments are conducted on benchmark functions to evaluate the performance.