Editorial: Special issue on advances in swarm intelligence for neural networks

Very recently, swarm intelligence (SI) became the hottest paradigm in the community of computational intelligence and has received extensive attention from many researchers. As we know, SI is the collective problem-solving behavior of groups of animals or artificial agents that result from the local interactions of the individuals with each other and with their environment. As usual, SI systems are primarily inspired by natural systems and greatly depend on certain key principles such as decentralization, stigmergy, collaboration, and self-organization which are observed in the organization of social insect colonies and other animal aggregates, such as ant colonies, bird flocks, fish schools, bacterial foraging, honey bee, fireworks explosion, brainstorming process, etc. Besides the research on theoretical analysis and algorithms, extensive application research of SI have also been carried out, in particular, the swarm intelligence for neural networks. Up to now, there are a number of research articles to deal with the applications of SI in neural networks which would inspire certain new research directions and solutions in the community of neural networks as well as swarm intelligence. The annual international conference on swarm intelligence (ICSI) (official website: http:// www.ic-si.org) eventually became one of the most important forums for scientists, engineers, educators, and practitioners to exchange the latest advantages in theories, technologies, and applications of swarm intelligence and related areas and attracts hundreds of researchers all over the world each year. This special issue included 24 highly evaluated papers from the third international conference on swarm intelligence (ICSI) (http:// www.ic-si.org), which was held from June 15 to 18, 2012, Shenzhen, China. All the papers were thoroughly revised and have been extended essentially by authors and then re-submitted to Neurocomputing for a regular peer reviewing process. These papers are divided into 5 groups which are briefly summarized as follows. Papers in the first group is to present improvements and theoretical analyses of some typical swarm intelligence algorithms including the genetic algorithm (GA), particle swarm optimization (PSO) and ant colony optimization (ACO). In particular, some of the improvements were specialized for certain type of applications. Zhang et al. proposed a new fitness scaling method, named powered distance sums scaling (PDSS), to eliminate the influence of fitness distribution on stochastic selection. PDSS maintains a more constant and consistent selective pressure in different optimization conditions and may help GA designers in balancing exploration and exploitation during evolution procedures. Hsu et al. proposed a genetic algorithm to solve MEDP. In comparison to the multi-start simple greedy algorithm and the ant colony optimization method, the proposed GA method performs better in most of the instances in terms of solution quality and time. Cheng et al. proposed an improved multi-objective particle swarm optimization with preference strategy (IMPSO-PS) for optimal integration of distributed generation (DG). The method introduces the dynamic selection of the global bests and a special mutation operation. The results show that the proposed approach can provide a wider range of Pareto solutions of high quality while satisfying special preference demands. Yusoff et al. proposed an improved discrete particle swarm optimization (DPSO) algorithm for solving the evacuation vehicle assignment problem (EVAP). The results show that DPSO with a min–max approach offers a good performance with respect to maximizing the number of individuals who can be evacuated by vehicles. Vitorino et al. presented a mechanism based on the artificial bee colony to generate diversity when all particles of the PSO converge to a single point of the search space. The improved method is named as ABeePSO, which is evaluated and compared to other well-known swarm based approaches using many benchmark functions. Deng et al. proposed an improved ant colony optimization (ACO) algorithm called pheromone mark ACO (PM-ACO) for the non-ergodic optimal problems. The PM-ACO associates the pheromone to nodes and has a pheromone trace of scattered points referred to as pheromone marks. Experimental results show that the improved PM-ACO has a good performance when applied to the shortest path problem. The four papers in the second group are to consider various emerging swarm intelligence algorithms inspired from nature, such as bacteria foraging optimization (BFO), artificial bee colony optimization (BCO) and firework algorithm (FWA). Applications and improvements of these algorithms are also included. To achieve high-quality solutions for constrained optimization problems, Niu et al. employed two modified bacterial foraging optimization (BFO), BFO with linear and non-linear decreasing chemotaxis step (BFO-LDC and BFO-NDC) to balance global search and local search. Wu et al. proposed a new model based on an assumption that the plasmodium of Physarum polycephalum forages for food along the gradient of chemo-attractants on a nutrient-poor substrate. Growth of Physarum is determined by the simple particle concentration field relating the distance to food source and the shape of food source. The model can imitate Physarum to avoid repellents and performs well in spanning tree construction. Maeda et al. presented a reduction of the artificial bee colony algorithm for global optimization. Bees sequentially reduce to reach a predetermined number of them grounded in the evaluation value. The proposed method had superiority in comparison with existing algorithms for complicated functions.