Under-informed momentum in PSO
Christopher K. Monson, Kevin D. Seppi · 2014
Particle Swarm Optimization is fundamentally a stochastic algorithm, where each particle takes into account noisy information from its own history as well as that of its neighborhood. Though basic information-theoretic principles would suggest that less noise indicates greater certainty, the momentum term is simultaneously the least directly-informed and the most deterministically applied. This dichotomy suggests that the typically confident treatment of momentum is misplaced, and that swarm performance can benefit from better-motivated processes that obviate momentum entirely.