Complex network analysis of firefly algorithm population dynamics
Magdalena Metlická, Donald David Davendra · 2017
This paper succeeds former work on complex networks formed by evolutionary algorithms dynamics, offering the insight into communication model of yet another type of swarm based algorithm, the Firefly Algorithm. FA's structure differs from that of Artificial Bee Colony Algorithm or Differential Evolution explored in preceding works, consequently also the population exhibits different communication patterns, posing a challenge to applying complex network analysis meaningfully. Three versions of Adaptive FA algorithm based on complex network analysis are presented, using the information obtained from vertex centralities to steer the population development in order to prevent premature convergence and stagnation. The experimentation conducted on 15 selected difficult multimodal optimisation problems shows that the FA benefits from utilization of complex network analysis derived information to augment the population dynamics.