Singular value relation for an LMI

82 Views Asked by At

Can we have the following?

For matrices $A$ and $B$, if $A \succeq B \implies \overline\sigma(A) \ge \overline\sigma(B)$?

where $\overline\sigma(\cdot)$ means the largest singular value.

1

There are 1 best solutions below

0
On

Yes" if B⪰0

and "no" otherwise.

When A,B⪰0 , singular values coincide with eigenvalues. By the variational characterisation of eigenvalues, we have λi(A)≥λi(B) in general. So, in particular, we have σmax(A)=λmax(A)≥λmax(B)=σmax(B)

.

If B is not positive semidefinite, the inequality doesn't hold. A simple counterexample is given by A=0⪰−I=B, where σmax(A)=0<1=σmax(B).