Article Preview
TopIntroduction
Particle swarm optimization (PSO) is a stochastic global optimization technique inspired by social behavior of bird flocking or fish schooling. In the conventional PSO suggested in Kennedy and Eberhart (1995) and Eberhart and Kennedy (1995), each particle in a population adjusts its position in the search space according to the best position it has found so far, and the position of the known best-fit particle in the entire population. Compared to other population-based algorithms, i.e., genetic algorithms, the PSO does not need genetic operators such as crossover and mutation. Thus it has advantages of easy implementation, fewer parameters to be adjusted, strong capability to escape from local optima as well as rapid convergence. As a result, the PSO outperforms other population-based algorithms in many real-world application domains.
In recent years, the PSO has been increasingly used as an efficient technique for solving complicated and hard optimization problems, such as function optimization, evolving artificial neural networks, fuzzy system control, optimization in dynamic and noisy environments, blind source separation, machine learning, games, to name a few. Furthermore, the PSO has also been found to be robust and fast in solving non-linear, non-differentiable and multi-modal problems (Ge & Zhou, 2005). Therefore, it is very important and necessary to exploit some new mechanisms and principles to improve and promote the performance of the conventional PSO for a variety of problems in practice. In this article, the clonal mechanism found in natural immune system of creatures is introduced into the PSO, resulting in a novel clonal PSO (CPSO, for short). In addition, in order to improve the CPSO further, an advance-and-retreat(AR) strategy and the concept of random black hole(RBH) are then introduced into the CPSO, resulting in two variants of the CPSO, called CPSO with AR strategy (AR-CPSO, for short) and RBH model (RBH-PSO, for short).
This article is an extended version of our earlier short paper (Tan & Xiao, 2007), in which a basic idea of the CPSO is briefly presented. Here, we have extended it substantially and included two variants with some deep discussions, comprehensive experimental studies as well as our application to spam detection.
The remainder of this article is organized as follows. Section II describes the conventional PSO algorithm and its related modification versions. Section III presents the proposed CPSO by introducing the clonal mechanism in NIS into the conventional PSO and its implementation. Section IV improves the CPSO by introducing the AR strategy and the RBH model. Section V gives several experimental results to illustrate the effectiveness and efficiency of the proposed algorithms in comparison with the conventional PSO. An application of spam detection is also given in details in section VI. Finally, concluding remarks are drawn in Section VII.