Skip to main content
12 events
when toggle format what by license comment
Aug 3 at 18:21 comment added Dima Pasechnik perhaps we are mixing up two types of similarity - od matrices, and of quadratic forms.
Aug 3 at 18:02 vote accept Aditya Bandekar
Aug 3 at 17:56 comment added Branimir Ćaćić This argument is correct. Indeed, since $B$ is assumed to the symmetric positive definite, it admits a unique symmetric positive definite square root $\sqrt{B}$ that you can construct via any orthogonal diagonalization of $B$; hence, take $g := \sqrt{B}$. @DimaPasechnik in general, $I = g^{-1} B g^{-t}$ need not be similar to $B$ precisely because $g$ need not be orthogonal.
Aug 3 at 17:16 comment added Aditya Bandekar This looks correct. I'll think about it for a while and convince myself before accepting the answer.
Aug 3 at 17:11 comment added Joseph Thanks for your comments. I have edited my answer to include more detail.
Aug 3 at 17:10 history edited Joseph CC BY-SA 4.0
added 638 characters in body
Aug 3 at 16:56 comment added Dima Pasechnik the best you can assume is that B is diagonal.
Aug 3 at 16:55 comment added Dima Pasechnik B=I makes no sense - as any matrix conjugate to I equals I.
Aug 3 at 16:23 comment added Aditya Bandekar He knows $B=gg^T=\frac{A+A^T}{2}$ is symmetric positive-definite, because he wrote $A$ in terms of its symmetric and antisymmetric parts. So $B=\frac{A+A^T}{2}$ and $C=\frac{A-A^T}{2}$. The assumption given in the question is $B$ is positive-definite. I'm stuck on the line "Thus, by acting by this group, we can assume $B=1$". The rest of the proof I understand. Can someone make this more clear for me?
Aug 3 at 16:19 comment added Dima Pasechnik Only symmetric p.d. matrices are of the form $gg^\top$. So your argument breaks down
Aug 3 at 15:47 history edited Joseph CC BY-SA 4.0
edited body
Aug 3 at 15:40 history answered Joseph CC BY-SA 4.0