Study On Population Evolutionary Mechanism Of Estimation Of Distribution Algorithms

Although some progress has been made since the birth of estimation ofdistribution algorithms,there are many issues to be studied such as algorithm theoreticalanalysis on the mechanism of evolution and population, algorithm design andapplication and so on. Population evolutionary of estimation of distribution algorithmsmechanisms mainly includes algorithm mode theorem and interconnections between thebiodiversity and the convergence of the algorithm.As the interconnections betweenvariables is the nature of the algorithm, which determines the difficulty of algorithmsfor solving problems.Algorithm of biodiversity for the estimation of distributionalgorithm to avoid a local optimum is not only of great significance,but also is a veryimportant problem in the theory of estimation of distribution algorithm. To improve theefficiency,Convergence is one of the most important indicator of estimation ofdistribution algorithm,determining whether the algorithm can search correctly andeffectively or not,in solving problems in the past which can not be resolved oreffectively.The paper presents convergence and diversity of the algorithm based on estimationof distribution algorithms,combined with other intelligent algorithms.The content of thispaper is as follows:(1) This paper mainly introduces a summary description of the estimation ofdistribution algorithm, led to a population estimation of distribution algorithmsevolutionary mechanism of research achievements and shortcomings,finally concludeswith the direction and methods.(2) Focuses on the frame structures estimation of distribution algorithm for,theresearch and development of the discrete and continuous of the estimation ofdistribution algorithm,some basic theories of population evolutionary mechanism andestimation of distribution algorithm, laying the Foundation for later study.(3) Research on distribution estimated algorithm in the of population evolutionmechanism, especially on convergence of distribution estimated algorithm for theinfluence of the algorithm, briefly analysis Ant Colony algorithm, then presents a fusionof algorithm based on Ant Colony algorithm and distribution estimated algorithm,proved by random process theory in the of Markov chain on the algorithm.Theexperimental simulation indicates that new algorithm is some convergence. (4) Through the analysis of species diversity in the estimation of distributionalgorithm, introduced a uniform mutation,and combined with the univariate marginaldistribution algorithm to propose a way to maintain diversity algorithm for univariatemarginal distribution algorithm, showing the effectiveness of new algorithm inmaintaining species diversity，via the experiments comparison.