A performance study on the effects of noise and evaporation in Particle Swarm Optimization

This paper presents a performance study on the effects of noise and evaporation in two variants of Particle Swarm Optimization (PSO) on large-scale optimization problems. The variants in consideration are the Synchronous PSO (S-PSO) and the Random Asynchronous PSO (RA-PSO), both of which are evaluated upon the set of benchmark functions presented at the IEEE CEC´2010 Special Session and Competition on Large-Scale Global Optimization. Results show an important detriment to the performance of both variants in the presence of different levels of noise. However, such detriment is significantly mitigated by incorporating an evaporation mechanism into particles to deal with such disruptive effects. Moreover, results show that RA-PSO is significantly better than S-PSO, more tolerant to noise, and better suited for the evaporation mechanism.