Matrix inequality with convex elements

67 Views Asked by At

I have the following constraint: $$ \begin{bmatrix} t & f(x) \\ g(y) & 1 \end{bmatrix} \succ 0$$

where $f(x),g(y)$ are convex (real) positive functions of $x$ and $y$, and $t\in \mathbb{R}_+$. Is this type of constraint convex and solvable in an interface like YALMIP or CVX? Can it be transformed into an LMI?

Thanks!

1

There are 1 best solutions below

0
On BEST ANSWER

If you mean that this matrix is positive definite, then this constrain is not convex unless $f(x)$ and $g(y)$ are linear function.

For a 2-dimensional matrix, that it is positive definite is equivalent to: $$ t>0,1>0,f(x)=g(y), f(x)g(y)-t<0 $$ i.e.: $$ f(x)=g(y), f(x)g(y)-t<0 $$ If you want these contraints to be convex, then $f(x), g(y)$ should be linear and $f(x)g(y)$ should be convex.