A Complementary Cyber Swarm Algorithm

Abstract

A recent study (Yin et al., 2010) showed that combining particle swarm optimization (PSO) with the strategies of scatter search (SS) and path relinking (PR) produces a Cyber Swarm Algorithm that creates a more effective form of PSO than methods that do not incorporate such mechanisms. This paper proposes a Complementary Cyber Swarm Algorithm (C/CyberSA) that performs in the same league as the original Cyber Swarm Algorithm but adopts different sets of ideas from the tabu search (TS) and the SS/PR template. The C/CyberSA exploits the guidance information and restriction information produced in the history of swarm search and the manipulation of adaptive memory. Responsive strategies using long term memory and path relinking implementations are proposed that make use of critical events encountered in the search. Experimental results with a large set of challenging test functions show that the C/CyberSA outperforms two recently proposed swarm-based methods by finding more optimal solutions while simultaneously using a smaller number of function evaluations. The C/CyberSA approach further produces improvements comparable to those obtained by the original CyberSA in relation to the Standard PSO 2007 method (Clerc, 2008). were previously found as a basis for focusing the search in regions anticipated to harbor additional solutions of high quality. Diversification promotes the exploration of regions appreciably different from those previously examined in order to produce new solutions with characteristics that depart from those already seen. Intensification and diversification work together to identify new promising regions when the slave heuristics stagnate in the executed search courses. Many intelligent algorithms fall in the territory of metaheuristics. served as a program committee member in many international conferences. He has also edited two books in the pattern recognition area. His current research interests include artificial intelligence, evolutionary computation, metaheuristics, pattern recognition, machine learning, and operations research. Fred Glover is the Chief Technology Officer in charge of algorithmic design and strategic planning initiatives for OptTek Systems, Inc., heading the development of commercial computer software systems currently serving more than 80,000 users in the United States and abroad. He also holds the title of Distinguished Professor at the University of Colorado, Boulder, where he is affiliated with the Leeds School of Business and the Department of Electrical and Computer Engineering. Dr. Glover is widely known for his work in the applications of computer decision support systems, including industrial planning, financial analysis, systems design, energy and natural resources planning, logistics, transportation and large-scale allocation models. He has authored or co-authored more than 400 published articles and eight books in the fields of mathematical optimization, computer science and artificial intelligence. He is also the originator of Tabu Search (Adaptive Memory Program...

A Complementary Cyber Swarm Algorithm

“…The Cyber Swarm Algorithm (CyberSA) of Yin et al (2010) creates an enhanced form of swarm algorithms by incorporating three features: (1) augmenting the information sharing among particles by learning from the reference set members, (2) systematically generating dynamic social networks in order to choose various solutions as the leaders such that the search can adapt to different functional landscape, and (3) executing diversification strategies based on path relinking approaches as a response to the status of the adaptive memory. The success of this method has motivated us to examine another variant that draws on alternative ideas from the same sources.…”

“…As previously noted, the choice of neighborhood topologies and leading solutions significantly affects the particle swarm performance. The literature discloses that the use of a dynamic neighborhood (Miranda et al, 2007;Clerc, 2008;Yin et al, 2010) and the local best solution lbest (Kennedy, 1999;Clerc, 2008) leads to a better performance. These notions create a form of multiple neighborhood search in which the neighboring particles (each maintaining a search trajectory) are selected at random or systematically and the local optimum corresponds to the best solution encountered by the multiple search trajectories.…”

“…We employ the policy widely adopted in the literature of representing the efficiency measure as the mean number of function evaluations required by a given algorithm in order to obtain an objective value that is sufficiently close to the known global optimum by reference to a specified gap. Yin et al (2010) have shown the advantages of the original CyberSA by comparing it with several other metaheuristics such as the Standard PSO 2007 (Clerc, 2008), C-GRASP (Hirsch et al, 2007), Direct Tabu Search (Hedar & Fukushima, 2006), Scatter Search (Laguna & Marti, 2005), and Hybrid Scatter Tabu Search (Duarte et al, 2011a). In our present comparison we also include the PSwarm algorithm of Vaz and Vicente (2007), which embeds the swarm algorithm into the pattern search framework.…”

“…In additional to the original mesh search, the particle swarm search can explore the nonconvexity of the objective function. More recently, Yin et al (2010) introduced a Cyber Swarm Algorithm (CyberSA) as the marriage of PSO and the scatter search/path relinking (SS/PR) template that obtains improved outcomes. With the addition of an external memory, embodied in a reference set consisting of the best solutions observed throughout the evolution history, useful information is produced and maintained that is not attainable by relying on traditional PSO mechanisms involving particle experiences.…”

“…

A recent study (Yin et al, 2010) showed that combining particle swarm optimization (PSO) with the strategies of scatter search (SS) and path relinking (PR) produces a Cyber Swarm Algorithm that creates a more effective form of PSO than methods that do not incorporate such mechanisms. This paper proposes a Complementary Cyber Swarm Algorithm (C/CyberSA) that performs in the same league as the original Cyber Swarm Algorithm but adopts different sets of ideas from the tabu search (TS) and the SS/PR template.

…”