Study On Swarm Intelligence And Their Applications

With high-speed development of society and technique,optimization problems exist in many applicable fields.However,most of them do not possess continuous or differentiable conditions.Therefore,other optimization algorithms,apart from traditional mathematical algorithms,should be explored.This paper focuses on intelligent optimization algorithms,makes compensate to their drawbacks and improves their performance,and finally applies them to real world applications.Among these evolutionary algorithms,particle swarm optimization has a rapid convergence speed;differential evolution exhibits better global search ability yet slower convergence speed;artificial bee colony holds the best global search ability but poorest local search capability.Drawing these concerning issues,this paper utilizes PSO and DE's merits and makes improvement on ABC algorithm,and finally applies them to real world applications.The main contributions of this paper are:This paper first analyzes PSO,DE and ABC's search equations,extracts PSO and DE's merits,and applies them to ABC in the employed bees' phase to accelerate the convergence speed.Besides,the concept of global best position is introduced to the onlooker bees' search equation in ABC to enhance the fine tuning capability.For scout bee's search mechanism,this paper reports a threshold-based activating strategy.Along with a new search equation,scouts are supposed to make local search around a food source.To test the proposed algorithm's optimization performance,a set of benchmark functions are applied.Secondly,this paper introduces PSO's merits to artificial bee colony algorithm.Two novel search equations are proposed for employed bees and onlooker bees for the purpose of accelerating the convergence speed and enhancing the solution accuracy.In addition,a new mutation strategy for scouts is proposed to further maintain the global search ability.Finally,the enhanced algorithm is applied to the multi-level thresholding image segmentation problem to demonstrate its effectiveness on real world applications.Thirdly,concerning the slow convergence speed of artificial bee colony algorithm,this paper proposed intelligent learning strategy for employed bees and onlooker bees.In the employed bees phase,a new turbulent operator is proposed to make better balance between global and local search.Besides,a new direction-based search strategy is proposed to avoid the oscillation phenomenon in employed bees' phase.In addition,an intelligent learning strategy for the worst employed bee is also proposed to enhance the solution accuracy.To test the proposed algorithm's efficiency,the multi-level thresholding image segmentation problem is applied.Finally,this paper puts forward an intelligent learning strategy for employed bees by further studying PSO,DE,and ABC's search strategy.Besides,a novel dimension selection strategy,from one-dimension to total-dimension,is proposed,aiming at maintaining global search ability at the initial stage of optimization and enhancing solution accuracy at the later stage of optimization.Finally,this proposed algorithm is applied to the feature-selection problem to demonstrate its effectiveness.