2
$\begingroup$

Let A be a linear transformation matrix f(x)=Ax, and if D is A in the eigenvector basis, $$D = Q^{-1}AQ$$, where Q is eigenvector matrix of A and a change of basis matrix changing A in standard basis to a basis of eigenvectors of A. It seems to be similar to eigendecomposition $$ A = Q\Lambda Q^{-1} $$ $$ \Lambda = Q^{-1}AQ$$ So the linear transformation matrix D with eigenvectors of A as basis is just the eigenvalues matrix of A. What are the connections here?

$\endgroup$
  • $\begingroup$ Other than whether the diagonal matrix is called $D$ or $\Lambda$ there seems to be no difference between your first and last equations. $\endgroup$ – Henning Makholm Jan 23 '18 at 20:17
1
$\begingroup$

You simply have different cases of the same result:

If $A$ is the matrix that represents a linear transformation in the standard basis and $Q$ is a matrix that has as columns the vectors of a new basis, than the same linear transformation is represented in this new basis by the matrix $B= Q^{-1}AQ $. This result is always true, for any matrix $A$.

Obviously we have also: $A=QBQ^{-1}$ and the eigendecomposition is a special case.

If $A$ is diagonalizable than it has a set of eigenvectors that forms a basis and, in this basis, the linear transformation is represented by a diagonal matrix $D$ that has as diagonal elements the eigenvalues. In this case the transformation matrix $Q$ has as columns the eigenvalues and, using the same formula, we have that $D=Q^{-1}AQ$

Also if the linear transformation is not diagonalizable than we can represent it in a canonical form (the Jordan canonical form) in a basis that is formed by the proper and the generalized eigenvectors. In this basis the transformation is represented by a matrix $\Lambda$ that is given by the same formula $\Lambda=Q^{-1}AQ$ (and $A=Q\Lambda Q^{-1}$) where, now, $Q$ is the matrix that has as columns the proper and generalized eigenvectors of $A$.

$\endgroup$
  • $\begingroup$ Thanks. If the basis is not eigenvectors, then is D still a diagonal matrix? (It is diagonal with eigenvalues when the eigenvectors are basis.) $\endgroup$ – drerD Jan 31 '18 at 8:11
  • 1
    $\begingroup$ No. The transformation is represented by a diagonal matrix only in the eigenbasis. $\endgroup$ – Emilio Novati Jan 31 '18 at 8:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.