Design and Analysis of Adaptive Swam Intelligence Approach for Multimodal Function Optimization in Bioinformatics

Abstract

Swarm Intelligence based algorithms have shown to have much promise, in a wide variety of complex newlineoptimization problems. These algorithms are found to be very efficient, robust and also very simple to newlineimplement. newlineOptimization has been an active research field for several past decades. Although there is a rich newlineassortment of established algorithms, but intrinsically they are very complex due to their mathematical newlinecharacteristics. This therefore has necessitated a categorization of algorithms, specific to the problem newlinedomain, sharing some common properties. Furthermore, different instances of the same problem may newlinehave different computational requirements. Hence it has always been a challenge and scope to the newlineresearchers to innovate a better technique. newlineParticle Swarm Intelligence (PSO) is one such emerging area, within the computational swarm newlineintelligence, which is inspired from bird flocking mechanism. Due to its simplicity, ease of newlineimplementation, effectiveness and parallel computation, PSO is an ideal tool for solving optimization newlineproblems. Especially it is a boon for the complicated problems that are considered to be impractical to be newlinesolved by traditional methods. newlineAs part of this research effort our attention is exclusively on PSO algorithm and its capability to tackle newlinemultimodal function optimization problems. Despite of its efficiency and capability to locate the good newlinesolution at a significantly faster rate as compared to other evolutionary optimization methods, PSO has a newlinetendency to get trapped in a local optimum. This premature convergence makes it difficult to find global newlineoptimum solutions for multimodal problems. Thus, there is a strong need to have a proper balance newlinebetween exploration and exploitation. Another drawback of conventional PSO is, its comparatively weak newlineability to fine tune the optimum solution. This is because of lack of diversity at the end of the search. To newlinefurther add, conventional PSO, do not consider variation of the objective function which also leads to newlinepremature convergence.