Skip to main content
5 events
when toggle format what by license comment
Jun 13, 2020 at 12:21 comment added dohmatob Your problem can be reduced to that of computing the quantity $z(A,B) := \mathbb E_{\theta \sim \tau}[\|\theta\|_A\|\theta\|_B]$. This is because the other terms have trivial formulae, since they are expectations of quadratic forms of a random vector whose covariance matrix you know explicitly. I don't think you can compute $z(A,B)$ in closed-form for general $A$ and $B$. You can definitely obtain (possibly crude!) lower and upper bounds for $z(A,B)$ in terms of the eigenvalues of $A$ and $B$.
S Feb 29, 2020 at 12:07 history suggested Louis Deaett
added tag linear-algebra
Feb 29, 2020 at 1:49 review Suggested edits
S Feb 29, 2020 at 12:07
Feb 27, 2020 at 18:31 history edited YCor
edited tags
Feb 27, 2020 at 17:47 history asked Lénaïc Chizat CC BY-SA 4.0