三維數(shù)據(jù)分析與模式識(shí)別方法及其在藥學(xué)研究中的應(yīng)用.pdf

1、湖南大學(xué)博士學(xué)位論文三維數(shù)據(jù)分析與模式識(shí)別方法及其在藥學(xué)研究中的應(yīng)用姓名：謝洪平申請(qǐng)學(xué)位級(jí)別：博士專業(yè)：分析化學(xué)指導(dǎo)教師：俞汝勤沈國(guó)勵(lì)2002.6.1ABSTRACTABSTRACTTheauthorofthepresentthesishasstudiedsomenewapproachesofchemicalrankestimationofathreewaydataarrayandtheapplicationofthreewaydata

2、analysisandchemicalpattemrecognitioninpharmacologyThemainresultsaresummarizedaSfollows1Thechemicalrankestimationofathreewaydataarray(ChaptersJoinPartJ)：Anunfoldedmatrixwiththerowspaceoffullrankisformedbyunfoldingathree—w

3、aydataarrayalongitsfullrankmodeWhenanorthogonalprojectioniscarriedoutfortheunfoldedmatrixusinganorthogonalprojectivematrixconstructedbyavectorwiththemaxinlumFrobeniusnorminthecolumnspaceoftheunfoldedmatrix，whichiscalledt

4、heprincipalnorrflvectorthemathematicrankoftheresidualmatrixshoulddecreaseonecomparedwiththatoftheunfoldedmatrixBasedOntheaforementionedfacts，aprincipalnolmvectororthogonMprojection(PNVOP)approachwasproposedtoestimatethec

5、hemicalrankofathreewaydataarrayTheexperimentsshowthatthemethodisrobustinresistingtheheteroscedasticnoiseThePNVOPmethodisadaptabletoestimatingthechemicalrankofathreewayarraywithafullrankmodeIfthetwoprincipalcomponentsuhsp

6、aceswiththeSalliesizeinthesainemodespacesofthetwounfoldedmatrixesformedbyunfoldingathreewayarrayalongitstwofullrankmodesallaresignalspaces，theirdifferencesshouldbeverysmallOtherwise，thesedifferencesshouldbesignificantBas

7、edonthisfhcLatwomodesubspacecomparison(TMSC)approachisproposedtoesfimatethechemicalrankofathreewayarrayItisrobusttoaveryhiglldegreeofcollinearitybetweenthespectraorchromatogramsinvolved，ortoaveryhighlevelofnoisecontained

8、inathreewayarraythoughcomparingtoPNVOPmethod，themethodrequiresthatthreearetwofullrankmodesforathree—waydataarray2Theapplicationsofthreewaydataanalysisandchemicalpatternrecognitioninpharmacologyresearch(Chapters36inPart2)