您好,欢迎访问三七文档
当前位置:首页 > 建筑/环境 > 电气安装工程 > 4补充关于矩阵的知识.
主要内容问题的引入秩亏自由网平差的原理广义逆的补充知识秩亏自由网平差的解法秩亏自由网平差解的性质补充一:六种常用的矩阵迭代法解方程组;高斯消去法;矩阵的三角分解(LU):任一非奇异矩阵可分解为:A=LU,其中一个是单位三角矩阵一、三角形矩阵LUfactorization.[L,U]=LU(A)storesanuppertriangularmatrixinUandapsychologicallylowertriangularmatrix(i.e.aproductoflowertriangularandpermutationmatrices)inL,sothatA=L*U.Acanberectangular.[L,U,P]=LU(A)returnsunitlowertriangularmatrixL,uppertriangularmatrixU,andpermutationmatrixPsothatP*A=L*U.补充一:六种常用的矩阵一定是方阵特征值均为实数特征向量正交其逆也对称二、对称矩阵DQPBBNNccPBBNbbAAPNaaTaaTT、、11三、正定矩阵与非负定矩阵性质:四、正交矩阵0)det(,0)4(0,0,)3(,)2(G)1(2AAAxxxRxABBBGATnT的所有主子式、有、、列满秩,、满秩对称阵定义:特征值全部为正(或非负)的实对称矩阵定义:设n×n阶方阵A,每行每列元素的平方和为1,每两列相应相应元素的积之和为0。也是正交阵、是正交阵,、、、正交阵未必是对称阵也是正交阵是正交阵,、、、、非奇异,且BA)4()3()2(1)det()1(1-ABBAAAAIAAAATTTcossinsincosA五、幂等矩阵六、初等矩阵AAA2,满足定义:方阵是方阵,分为三类:nmmBRARCRCBAIBArnAIRrARAIAIAtrARnnmmnnm.)()()(,5)(,)(4)(3)()(21012且为幂等则)设(则)设()()(即)特征值非性质:(加某一行倍数与另一行相某一行的倍数行交换单位阵其逆为同类型初等阵初等阵非奇异性质补充二:矩阵的六类数字特征一、秩(rank)定义:有一个k阶子式不为零,所有的k+1阶子式都为零))()(dim()()();()()(5)(,)(,)(4)()(3))(),...(),(min()...(2)()()()(12121TTnttmmmTTTBRARBRARBARBRARBARtABRtBRtARARQARQARARARAAARARAARAARAR)(则)(为满秩方阵,则)()()性质:(Matlab命令:rank(A).RANK(A)providesanestimateofthenumberoflinearlyindependentrowsorcolumnsofamatrixA.二、行列式(determinant))det()det()det()det()det()det(,2),det()det()det(12112212112222121112122111122211211AAAAAAAAAAAAAAAAAAABABAABnn可逆,当可逆,当)(为同阶方阵、)性质:(思考1:如何证明?Matlab命令:det(A).DET(X)isthedeterminantofthesquarematrixX.UseCONDinsteadofDETtotestformatrixsingularity.三、迹(trace))()(1BAtrABtr)性质:(思考2:niiiaAtrA1)(,定义:方阵?MSE,)(,)(,22122221112211,1,)(求已知:XXDXEXnnnnnnnMatlab命令:trace(A)TRACE(A)isthesumofthediagonalelementsofA,whichisalsothesumoftheeigenvaluesofA.四、范数(Norm)范数、无穷范数、分类:二范数、一范数P,相容性(四个条件)矩阵:等式负性、齐次性、三角不向量:(三个条件)非......Formatrices...NORM(X)isthe2-normofX.NORM(X,2)isthesameasNORM(X).NORM(X,1)isthe1-normofX.NORM(X,inf)istheinfinitynormofX.NORM(X,'fro')istheFrobeniusnormofX.NORM(X,P)isavailableformatrixXonlyifPis1,2,infor'fro'.Forvectors...NORM(V,P)=sum(abs(V).^P)^(1/P).NORM(V)=norm(V,2).NORM(V,inf)=max(abs(V)).NORM(V,-inf)=min(abs(V)).五、特征值(Eigenvalue);特征向量(Eigenvector)xAxA,定义:方阵思考3:特征值与特征向量在求误差椭圆中的应用EIGEigenvaluesandeigenvectors.E=EIG(X)isavectorcontainingtheeigenvaluesofasquarematrixX.[V,D]=EIG(X)producesadiagonalmatrixDofeigenvaluesandafullmatrixVwhosecolumnsarethecorrespondingeigenvectorssothatX*V=V*D.六、奇异值(Singularvalue)的奇异值为,,,称设,个特征值为记其其特征值为非负,为对称阵,的矩阵,则为定义:设AnmknAAnmAnnT2,122221),,min(思考4:奇异值分解(SVD)与方程求解!SVDSingularvaluedecomposition.[U,S,V]=SVD(X)producesadiagonalmatrixS,ofthesamedimensionasXandwithnonnegativediagonalelementsindecreasingorder,andunitarymatricesUandVsothatX=U*S*V'.思考4:奇异值分解(SVD)与方程求解!121ˆˆˆ110011101321321XXXVVVU=-0.7071-0.40820.57740.7071-0.40820.5774-0.00000.81650.5774S=1.73210001.73210000.0000V=-0.8165-0.0000-0.57740.4082-0.7071-0.57740.40820.7071-0.5774(1)[U,S,V]=SVD(A)(2)Sp=pinv(S)Sp=0.57740000.57740000(3)X=(V*Sp*U')*b'X=-0.33331.0000-0.6667补充三:广义逆矩阵一、满秩长方阵的逆1、列满秩的逆:AmnTTLAAAA11IAAL1PAPAAATTL11-------左逆左逆不唯一:一般式:2、行满秩的逆:Amn11TTRAAAAIAAR111TTRAQAQAA-------右逆右逆不唯一:一般式:一、满秩长方阵的逆3、满秩长方阵逆的性质:1111111111)2()1(LRTRLTTTAAAAAAAAAAAA为行满秩,为列满秩,4、奇异单位矩阵:AAAAAAAARL110为行满秩,为列满秩,0AIAAL1注意与定义式的区别:IAAR1核心是矩阵的维数与奇异性二、广义逆(Generalinverse)1)降秩法A1、定义AmnAAAA满足A总存在,但不唯一右逆、左逆、凯莱逆均是广义逆(特殊逆)),min()(nmrAR2、广义逆的求法22211211))*(()*())*((*AAAArmrnrrnrmrnrrA000111AA1)满秩分解法2、广义逆的求法mrrnmnCBA*nrLrmRBCA11思考5:两种解法如何证明?上机实验:广义逆的解法满秩分解法的步骤rARAAArmnrnmn)(,211)*(**其中①1AB②令TTLBBBB11③求ABCL1④求11TTRCCCC⑤求nrLrmRBCA11⑥得3、广义逆的性质的广义逆也是的广义逆,则,为是幂等阵,且)(),()(正定,)若()(的常数,)设(其中之一)(AAGAAGARAARAAARARAAPAAPAAAPAAPAAAPAAAAAAAAAAAAAkAkAATTTTTTTTTTTTTTTTT)7()(6mnmin)()(5)(;)(4)(;)(31)k(02)()(1-三、伪逆(pseudoinverse)A1、定义AmnGAGAAGAGGGAGAAGATT)()4()()3()2()1(、、、、满足GA是唯一的),min()(nmrARMoore(1920年)和Penrose(1955年)提出!1)特殊的伪逆2、伪逆的计算1AnnaaaA2211例1:1LA1RA2)当A为对角阵时nnaaaA22110100iiiiiiiiaaaa当当402A25.005.0A3)一般矩阵的伪逆TTTTAAAAAAAA)()(4)当N为对称方阵GA是唯一的)(TAA)(AAT不唯一,但和NNNNNNNN)()(5)设AANTTANATAINA10)(lim11LRBCA6)设满秩分解,)()()(,,rCRBRARCBArmnrmn例2:101110011A211121112AANT633363336NN00002101291)(NN21112111291)()(NNNNNNNN11001110131TANAQROrthogonal-triangulardecomposition.[Q,R]=QR(A),whereAism-by-n,producesanm-by-nuppertriangularmatrixRandanm-by-munitarymatrixQsothatA=Q*R.3、伪逆的性质)()(3)(,,)()(2)(1ARARAAAAAAAATTTT)(则若)()(PINVPseudoinverse.X=PINV(A)producesamatrixXofthesamedimensionsasA'sothatA*X*A=A,X*A*X=XandA*XandX*AareHermitian.ThecomputationisbasedonSVD(A)andanysingularvalueslessthanatolerancearetreatedaszero.Thedefaulttolera
本文标题:4补充关于矩阵的知识.
链接地址:https://www.777doc.com/doc-2926873 .html