Research Of Improved Particle Swarm Optimization Using MapReduce

Increasingly complex real-life optimization problems bring new challenges to evolutionary computation.Distributed optimization algorithms have received widespread attention in the past decade as an effective mean to address this challenge.Map Reduce,as a representative platform of cloud computing,can be combined with optimization algorithms to promote the development of distributed evolutionary algorithms effectively.Particle swarm optimization algorithm is a classic evolutionary optimization algorithm,it has been concerned by scholars from all walks of life after it was put forward,and variety of improved algorithms are proposed.As classical ones of the improved algorithms,cooperative particle swarm optimization algorithm and quantum-behaved particle swarm optimization algorithm improve the performance of the original particle swarm optimization algorithm from different points.Based on the summary of the related fields,this paper proposes a new distributed model that is suitable for optimization algorithm based on the prototype of Map Reduce,and shows the performance of distributed algorithm through two improved particle swarm optimization algorithms.The specific content is as follows:(1)Brief introduce the research background of this paper,and put forward the real problems of big data and cloud computing,then lead to the distributed evolutionary algorithm.On the basis of the background,the research direction and achievements of evolutionary algorithm,particle swarm optimization algorithm,cloud platform and Map Reduce have been further introduced.And then over to the main content of this article and chapter arrangements.(2)Introduce the Map Reduce model in detail by programming mode,concrete realization,relative advantage and related content of Hadoop.On the basis of the basic model,this paper presents a new evolutionary algorithm model called MREA model.MREA is started from the feasible domain,divides the search space into a large number of non-overlapping subspaces.The Map function is evolved on the data block composed of subspaces,and the relative optimal solutions of each subspace from Map functions are delivered to the Reduce function.Reduce function is used to compare and select to get the global optimal solution.The proposed model is suitable for the vast majority of optimization algorithms with parallel potential.The transplant operation is simple and convenient.The solution to the large-scale complex problems can be greatly improved on the quality of the solution and efficiency of running time.(3)The Cooperative Particle Swarm Optimization Algorithm using Map Reduce is proposed.The cooperative particle swarm optimization algorithm is an evolutionary algorithm of dimension distribution.It divides the original high-dimensional population into multiple low-dimensional subpopulations,and avoids the "cures of dimensionality" to a certain extent by reducing the dimension of population,but it is still plagued by premature phenomenon.The original cooperative particle swarm optimization algorithm is transplanted onto the MREA model,which the proposed algorithm is not only reduced the dimension of the population,but also divided the domain of definition,and narrowed the search range of a single evolutionary algorithm."Double Simplification" has led to a significant improvement in the new parallel algorithm,both in terms of quality of solution and running time,and has demonstrated strong competitiveness compared to the CEC 2013 award-winning algorithms.(4)The Quantum-behaved Particle Swarm Optimization Algorithm using Map Reduce is proposed.The quantum-behaved particle swarm optimization algorithm moves the domain from the original classical space to the quantum space,making the search particles can be appeared anywhere in the domain.The wide distribution of the particles can overcome the premature phenomenon caused by the standard particle swarm optimization algorithm itself.However,with the expansion of the scale of the problems,the original serial algorithm has been a great restraint in the computing resources.In this paper,the quantum-behaved particle swarm optimization algorithm is transplanted onto the MREA model,which realizes the parallelization of the algorithm,greatly improves the efficiency of the algorithm,and greatly shortens the running time,and improves the quality of the algorithm.