基于多维数据分析的神经网络与分布式计算研究

《基于多维数据分析的神经网络与分布式计算研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、武汉理工大学博士学位论文基于多维数据分析的神经网络与分布式计算研究姓名：刘天桢申请学位级别：博士专业：计算机应用技术指导教师：童恒庆20080401武汉理工大学博士学位论文数据仓库与．NET技术开发，采用无监督学习的凸约束神经网络模型架构，实现了基于远程方法调用的分布式计算。I关键词J：多维数据分析，多维矩阵，人工神经网络，分布式计算，数据仓库Ⅱ武汉理工大学博士学位论文AbstractArtificialneuralnetworktechnologyisatopicconcernedbyscientistsinmanydomains，becauseofitscha

2、racteristicssuchasmassiveparallelprocess，distributedstorage，self-adaptability,fault-tolerantandSOon．Ithasbeenwidelyappliedinmanyfieldssuchasbiology,electronics，computerscience，mathematicsandSOon．WiththerapiddevelopmentofnetworkcommunicationtechnologyandInternet,thedistributedcomputing

3、hasbecomeoneofthekeytechnologiesinfluencingtoday’sdevelopmentincomputertechnology．Andithasbeenusedinmodernsocietyandeconomicdevelopment．Bothofthetechnologiesneeddata,however,lotsofdatacomefromthemultidimensionaldatastoredindatawarehouse．Bothofthetechnologiesneeddataanalysis，whichwilli

4、nvolvemultidimensionalmatrix．Therefore，ithasimportantmeaningtostudytheartificialneuralnetworksanddistributedcomputingbasedonmultidimensionaldataanalysis，SOOurresearchWassupportedbyNationalNaturalScienceFundofChin乱Thisdissertationisdividedintofourpartsasfollows．Thefirstpartfocusesonthe

5、studyofmultidimensionaldataanalysisandmultidimensionalmatrix．Weintroducetheconceptofmultidimensionalmatrix,accordingtonecessityofusingmultidimensionaldataanalysisindatawarehouse．Thenwediscussthepropertiesofcubicmatrixwhichhasthemostwidelyapplicationinmultidimensionalmatrix,SOweestabli

6、shbasisforapplicationinneuralnetworkanddistributedcomputing．Thesecondpartfocusesonthestudyofartificialneuralnetworksbasedonmultidimensionaldataanalysis．Atfirst,weproposesakindofunsupervisedlearningneuralnetworkmodel、析tllconvexconstraintwhichhasspecialstructureandcanrealizethecompressi

7、onofdataandreductionprocess．耶1emaincharacteristicsoftheneuralnetworkcanrepresentinformationafterbeingtrained．Secondly,westudyakindofBayesneuralnetworks，andadoptgeneralnaiveBayestohandlecontinuousvariables，then,proposeakindofkernelfunctionconstructedbyorthogonalpolynomialswhichisusedto

8、estim