A Peer-Peer Particle Swarm Optimizer

Particle Swarm Optimization is a parallel algorithm that spawns particles across a search space searching for an optimized solution. Though inherently parallel, they have distinct synchronizations points which stumbles attempts to create completely distributed versions of it. In this paper, we attempt to create a completely distributed peer-peer particle swarm optimization in a cluster of heterogeneous nodes. Since, the original algorithm requires explicit synchronization points we modified the algorithm in multiple ways to support a peer-peer system of nodes. We also modify certain aspect of the basic PSO algorithm and show how certain numerical problems can take advantage of the same thereby yielding fast convergence.