双对称线性矩阵方程的最佳逼近解（英文）.pdfVIP

数 学 杂 志 Vo1．31(2011) NO．2 J．ofMath．(PRC) THE AN TI—SYM M ETRIC oPTIM A L APPR0XIM ATIoN S0LUTIoN F0R A LINEAR MATRIX EQUATION PAIR LIN Hong—cheng (Dept．ofSciences，HunanfbreignEconomicRelationsandTradeCollege，Changsha410015，China) Abstract： Thisarticlediscussestheconsistentconditionsofthesymmetricsolution and symmetricleastsquaressolutionforthelinearmatrixequationpair(A xA，B B)： (C，D) obtainedbyChangandWang(1993)．Byusingtheprojectiontheorem inHilbertspace，thequo— tientsingularvaluedecomposition(QSVD)andthecanonicalcorrelationdecomposition(CCD)for efficienttoolstoobtaintheexplicitanalyticalexpressionoftheanti--symmetricoptimalapproxima—． tionsolution(ortheminimumFrobeniusnormanti—symmetricsolutionasaspecialcase)forthe least—squaresproblem ofthism atrix equationpair． Keywords： linearmatrixequation；leastsquareproblem ；optimalapproximationsolution； canonicalcorrelationdecomposition(CCD)；quotientsingularvaluedecomposition(QSVD) 2010MR SuhjectClassification： 15A06；15A18；15A24 Documentcode： A ArticleID： 0255．7797(2011)02—0233—12 1 Introduction Linearmatrix equations，leastsquaresproblemsandoptimalapproximation problems havebeenconsideredbyaseriesofliteratures(see[2]，[9]and[14])．Theseproblemshave beenofinterestinmanyat)I：ications．suchasinverseproblemsofvibrationtheory[4J．control theory，digitalimageandsignalprocessingjandmultidimensionalapproximation6【]． W esupposethatA standsforthetransposeofmatrixA．Rm ”isthesetofallm ×n realmatrices，SR凡×扎，AR礼×几 and ORn×n arethesetsofallsymmetric，anti—symmetric andorthogonalmatricesin R ．respectively．A JE7standsofrtheHadamardproductof matricesAandB (cf．[10])，andlly denotestheFrobeniusnormofarealmatrixY， definedas