Based On The Distribution Of Decomposition Of The Pair - Copulas Connect Estimation Algorithm Research

Estimation of Distribution Algorithm (EDA) is a new type of intelligentoptimization algorithm. It is developd based on binary code GeneticAlgorithm(GA).The difference between it and GA is that it uses the probabilitymodel of reflecting population to replace the crossover and mutation operation.The copula theory is a new theory in the field of statistics multivariatecorrelation. It has become a new research field based on the performancecharacteristics of the theory. It has now been used in the estimation of theprobability distribution model of advantage populations by many researchers.Because of the complexity of the high-dimensional copula theory, at presentmost of these researches on Estimation of Distribution Algorithm based on thecopula theory are used in low dimension cases. After carefully analyzing severalconstruction methods of high-dimensional copula---the general constructionmethod、the FNAC method、the HNAC method、 the PNAC method、thepair-copula decomposition method, this paper considers that the pair-copuladecomposition method has the structure characteristics of processing ahigh-dimensional copula to many two-dimensional copulas relying on differentvine structures step by step, and that copula function has the selective diversity.Therefore, in view of the constructing ability of the pair-copula decompositionmethod in high-dimensional correlation, this article proposes that Estimation ofDistribution Algorithm based on pair-copula decomposition method. The contentof this paper are mainly the following aspects:First, this paper briefly discusses the development background of EDAincluding the theoretical basis and the historical background and describes thealgorithm basic framework, the execution flow chart and the estimation ofdistribution algorithm based on the copula theory in detail.It analyzes theperformance of the algorithm and its advantages and disadvantages.Then, this paper detailily describes the copula theory and several kinds ofmethod to construct high dimensional copula，and also briefly expounds theadvantages and disadvantages of the pair-copula decomposition method in high-dimensional copula correlation analysis performance.Finally, This article mainly focuses on the estimation of distributionalgorithm based on pair-copula decomposition method, specifically expoundsthe basic theory that how to build a class of pair-copula function on takingadvantage of the pair-copula decomposition method（D-vine、C-vine）and howto use the condition pair-copula function to sample the probability model and thewhole process of the algorithm of the model. Lately this paper tells thealgorithm framework the estimation of distribution algorithm based onpair-copula decomposition method on the vine structure.The simulationexperiment of several typical benchmark functions with the algorithm andrelated algorithms is conducted, the experimental results show that the algorithmhas the better ability in solving high-dimensional scale optimization problemssignificantly.