2
$\begingroup$

For any $nxn$ positive definite symmetric matrix $A$ is it possible to write it's entries $a_{ij}$ as inner products of vectors $v_1,v_2,....,v_n$, that is $a_{ij}=\langle v_i,v_j\rangle$? Is there a deterministic way to find $v_1,v_2,....,v_n$ for any given $A$? I thought maybe Cholesky decomposition would be the best way to do it. Have I missed something? Is there any easier way?

Thanks!

$\endgroup$
2
$\begingroup$

Probably Cholesky decomposition is one of the best ways, as once we have written $A=BB^T$, whose row vectors are $v_1,\dots,v_n$ we are done. Conversely, if the $v_j$ such that $a_{ij}=\langle v_i,v_j\rangle$ are given, then we get the Cholesky decomposition writing $B:=\pmatrix{v_1^T\\ v_2^T\\\vdots \\ v_n^T}$.

$\endgroup$
1
$\begingroup$

"every positive semidefinite matrix is the Gramian matrix for some set of vectors.", see this.

$\endgroup$

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.