The global best (gbest) is the best candidate position found by an entire particle swarm so far. Unlike a personal best, which records one particle’s experience, gbest summarizes the swarm’s strongest result. When neighborhoods contain only some particles, their best is called the local best (lbest); when a neighborhood includes the whole swarm, its neighborhood best becomes gbest. section 9.2
PSO uses the global best to:
- Identify the strongest fitness result across the swarm after particles evaluate their current positions. section 9.2
- Attract particles toward collective experience while inertia and personal best information preserve exploration and individual learning. section 9.2.1
- Replace gbest with nbest or lbest when the swarm is divided into multiple neighborhoods. section 9.2.1
- Update either globally and simultaneously in synchronous PSO or from the latest available information in asynchronous PSO. section 9.2.2
Global-best reinforcement in ant colony system
In ant colony system, pheromone reinforcement is performed only by the ant with the global-best solution—the best path found so far. This elitist strategy can make the search converge too quickly around the current best path, so max–min ant system was developed to preserve exploration. section 10.3.3
Figure 9.2 The PSO algorithm
Use the PSO flowchart to follow initialization, fitness evaluation, best-position updates, particle movement, and termination. section 9.2
