How does SVD work if the $A^TA$ matrix is singular? Lets say that the matrix looks like $$\begin{matrix}X & 0 \\0 & 0 \\\end{matrix} $$ Where X is an arbitrary real valued number. Is it still possible to obtain $\Sigma$ ? or will it now be in Jordan form? What about the case of repeated Eignevalues? how will this now impact $\Sigma$ along with $U$ and $V$ Matrix?

  • 1
    $\begingroup$ Why do you think there is a problem when $A^\top A$ is singular? If this is the case, then some of the eigenvalues of $A^\top A$ are zero, so $\Sigma$ will have some zero diagonal entries. $\endgroup$
    – angryavian
    Commented Apr 1, 2018 at 18:26
  • 1
    $\begingroup$ $A^\top A$ is always symmetric. Thus, it is always diagonalizable. No need to bring Jordan forms to the discussion. $\endgroup$ Commented Jul 15, 2018 at 18:52
  • $\begingroup$ The Singular Value Decomposition is possible for any matrix. $\endgroup$
    – user65203
    Commented Jul 15, 2018 at 18:57

1 Answer 1


Singular value decomposition works the same whether A$^{\mathrm T}$A is singular or not or whether it has multiple eigenvalues or not. In all cases, $\Sigma$ will be diagonal (not necessarily of the form of the corresponding matrix in the Jordan decomposition).


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .