Parallel swarm optimization for web information retrieval

In this paper, we show that direct search methods are more suited and simpler to implement than the other search techniques for information retrieval. Two novel PSO algorithms are designed for the purpose of validating this important result. One of these algorithms is sequential and the other one is a parallel version. We discuss the advantages of these algorithms and demonstrate that not only they are suited for web information retrieval but they also outperform the existing algorithms from the design and experimental points of view. Extensive experiments were performed on CACM and RCV1 collections. Performances in terms of solution quality and runtime are compared between these algorithms and exact methods. Numerical results exhibit the superiority of parallel PSO on all the others in terms of scalability while yielding comparable quality.