矩阵数值特征和迹占优矩阵的分析-analysis of matrix numerical characteristics and trace dominance matrix.docx

摘要矩阵分析和矩阵计算在数学物理，计算数学，控制理论，图像处理，经济学等领域有着广泛的应用.本文主要研究了矩阵数值特征的估计和迹占优矩阵的性质及迭代算法，主要内容和创新点包括：21.对非正规矩阵秩的下界的估计进行了研究，得到了非正规矩阵秩的下界的三个估计，所得估计在很多情形下较KyFan-Hoffman不等式精确.22.对于h(A)??0其中h(A)??trA???n?1?A???AA*??A*A的矩阵得到其所有特征值都位于如下一个圆盘之中?trAFF2?trA??nh??A??????z?C:??z???nn??同时利用矩阵分块、相似变换和均值不等式得到了一般矩阵的包含所有特征值的圆盘定理.3.改进了经典的Brauer和KyFan定理，得到了如下特征值分布区域n???z?C:z??az??a?n1?????pa???n?1?????pa?i,j?1?ik?1????jl?1ppiijjkikljl?????i??j和n?k??i???l??j????z?C:z??aiii,j?1i??jz??ajj???。?B????bii??。??B???bii??aii?ajj??同时给出了矩阵张量积的特征值分布区域，并得到了特征值实部和虚部的几个估计.4.基于奇异值分解和不等式理论，得到了矩阵奇异值的上界和下界的估计，并给出了如下一个不等式2???2FB??A?Fdiagτ1??σ1,τ2?σ2,?,τn??σn5.基于迹占优矩阵和广义迹占优矩阵的概念，得到了迹占优矩阵和广义迹占优矩阵的一些性质；结合最优化理论和广义迹占优矩阵的性质，提出了判断一个矩阵是否是广义迹占优矩阵的算法并通过数值算例对所得算法进行了验证.关键词：秩，特征值，奇异值，迹占优矩阵，算法ABSTRACTMatrixanalysisandmatrixcomputationshavewideapplicationsinmathematicphysics,computationalmathematics,controltheory,imageanalysisandeconomicsetc.Thepurposeofthisthesisistodiscussthenumericalcharacteristicsofmatricesandthepropertiesanditerativealgorithmoftracedominantmatrices.Themainresultsandinnovationsareasfollows:Theestimationsforlowerboundsfortherankofnon-normalmatrixareinvestigated.Threelowerboundsfortherankareobtained.Inmanycases,the2estimationsaremoreaccuratethanKyFan-Hoffmaninequality.Forh(A)??0suchthath(A)??trA???n?1?A???AA*??A*A,weprove2FFthatalltheeigenvaluesarelocatedinthefollowingdisk???z?C:2trA??nh??A???trA?z???nn?????Basedonmatrixpartition,similaritytransformationandmeaninequalitythedisktheoremswhichconcludealltheeigenvaluesareshown.SomeimprovementsofBrauertheoremandKyFantheoremarepresented,theeigenvaluesinclusionregionsaregivenasfollowsn???z?C:z??az??a?n1p?????pa???n?p1?????pa?i,j?1?ik?1????jl?1andi??jniijjkikljl??????k??i???l??j????z?C:z??aiii,j?1i??jz??ajj???。?B????bii??。??B???bii??aii?ajj??Thedistributionsfortheeigenvaluesoftensorproductofmatricesareobtained,severalestimationsfortherealpartandimaginarypartofeigenvaluesareshown.Basedonsingularvaluedecompositionandinequalitytheories,theupperboundandlowerboundforthesingular