克莱夫·莫勒(Cleve Moler)是第一位Matlab的作者,foundersof MathWorks, and is currently Chief Mathematician at the company. He is the author of two books about MATLAB that are availableonline。他在这里写了关于MATLAB,科学计算和有趣的数学的文章。
Here is a static picture of the starting matrix. Notice that it is not symmetric.最初的减少用途N-2类似者类似的人一次将零以下方的圆柱引入圆柱。结果称为赫森伯格矩阵(不要让拼写检查员将其更改为Heisenberg matrix。)Now the QR algorithm gradually reduces most subdiagonal elements to roundoff level, so they can be set to zero. The corresponding diagonal element is an eigenvalue. The iteration count is shown in the title. The element below the diagonal in the last row is the initial target; it requires four iterations. The next two rows require three iterations each. The remaining subdiagonals require just one or two iterations each.All this is done with real arithmetic, although a real, nonsymmetric matrix may have complex eigenvalues. So the final matrix may have 2-by-2 bumps on the diagonal. This example has one bump in rows 3 and 4. The eigenvalues of a bump are a complex conjugate pair of eigenvalues of the input matrix. All the other diagonal elements are real eigenvalues of the input matrix.计算的特征值是:
Here is the static picture. (The computation is done on half of the matrix, but we show the entire array.)通过对称,将列为零的六个家庭也为零。现在,QR迭代仅在两个向量上起作用,即对角线和偏对角。极限是包含特征值的对角线。
Use a Householder operating from the left to zero a column and then another Householder operating from the right to zero most of a row.现在,双面QR迭代将对角线降低到可忽略的尺寸。所得的对角线包含奇异值。
Comments
要发表评论,请单击here登录您的数学帐户或创建一个新帐户。