For a matrix $A \in \mathbb R^{m\times n}$ with columns $a_i \in \mathbb R^m$, we know that $A^\top A$ is Positive Semidefinite (PSD).

I am interested in checking if B is positive semidefinite where $B_{ij} = (a_i^\top a_j)^2$. Each component of $B$ is the square of corresponding component of $A^\top A$.

  • $\begingroup$ if $A^\top A$ is diagonal it's obviiously trivial, as a feel I think that you can't conclude nothing in general, but I'm curious to see tha aswers to this question $\endgroup$ – gimusi Dec 5 '17 at 0:33

Yes. This is a direct consequence of the Schur product theorem: the elementwise product of two positive semidefinite matrices is also positive semidefinite.

  • $\begingroup$ The proof based on eigen decomposition is very illuminating. Thank you!. Was this very obvious to you or did you need to search for this ? $\endgroup$ – Hestenes Dec 5 '17 at 4:25
  • $\begingroup$ I had met this fact before, but I had forgotten its name, so it took a small bit of searching to find again. $\endgroup$ – kimchi lover Dec 5 '17 at 7:04

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.