LabVIEW MathScript RT Module

キャンセル
次の結果を表示 
次の代わりに検索 
もしかして: 

eigenvalue order

when I perform an eigen analysis mathscript gives the values as 0, 9, 0  (actually the first zero is -4.4e-16 and the second zero is 3.88e-32)  whereas matlab gives 0,0,9.  Is there a proper order in which the values are reported?
0 件の賞賛
メッセージ1/3
7,523件の閲覧回数
Hi,
There is no proper order for eigenvalues returned. It depends on your input matrix. If you want the eigenvalues in ascending order, you must sort them by yourself.
0 件の賞賛
メッセージ2/3
7,518件の閲覧回数
Hello,

Accurately determining when a result should be zero and when it is actually a very small number is difficult.  As ttrr mentioned, there is no proper ordering for eigenvalues.  If you choose to sort the eigenvalues yourself, remember that the eigenvectors correspond to a particular eigenvalue.  You will need to rearrange the columns of that matrix as well.

As for sorting in general, it is really a special case when the eigenvalues of a matrix are all real.  In this case, there are some obvious orderings; namely, ascending and descending.  However, in the more general case, eigenvalues are usually complex.  There is no truly intuitive way to sort complex numbers.  Sorting by magnitude is a logical choice, but then which value should be listed first: 3+4i or -3+4i?  What about 3+4i or 3-4i?

Could you provide the matrix you were looking at?  We could analyze the eigenvalue computation algorithm here and possibly improve it to return 0 instead of a small non-zero number.

Grant M.
Staff Software Engineer | LabVIEW Math & Signal Processing | National Instruments
0 件の賞賛
メッセージ3/3
7,486件の閲覧回数