Let $A\in\mathbb{R}^{n\times n}$ be a matrix with eigenvalues $\{\lambda_i\}_{i=1}^n$ such that $\mathrm{Re}(\lambda_i)<0$. Moreover, let $\succeq$ denote the standard partial order in the cone of positive definite matrices.

Consider the following maximization problem \begin{align}\label{eq:opt}\tag{$\ast$} M=\sup_{P,X\in\mathbb{R}^{n\times n}} \quad & \mathrm{tr}(XP^{-1})\\ \text{s.t.} \quad & AP+(AP)^\top=-X \\ & X=X^\top \succeq 0 \\ & \mathrm{tr}(X)=1 \\ & P=P^\top \succeq 0. \end{align}

Notice that we can rewrite \eqref{eq:opt} as \begin{align} M&=\sup_{\substack{X\in\mathbb{R}^{n\times n}\\ X=X^\top\succeq 0,\ \mathrm{tr}(X)=1}} -\mathrm{tr}((AP+PA^\top)P^{-1})=-2\mathrm{tr}(A). \end{align}

My question. If the cost function $\mathrm{tr}(XP^{-1})$ in \eqref{eq:opt} is replaced by $\mathrm{tr}(X(\alpha I+ P)^{-1})$, with $\alpha>0$, is it true that $$ M= \frac{1}{\alpha-\frac{1}{2\mathrm{tr}(A)}} \ \ \ ? $$

From an extensive amount of numerical simulations it seems that answer to my latter question is in the affirmative. However, after spending a lot of time thinking about this, I couldn't prove it. So, any help is really appreciated! Thank you!

  • What does $X>0$ mean? Could you properly add constraints to (*) and add P as an optimization variable? – LinAlg Oct 11 at 12:32
  • @LinAlg $X>0$ means that $X$ is positive definite. I’ve explicitly added the constraints to the problem. – Ludwig Oct 11 at 13:54
  • can you use the well known result that $(I+P)^{-1} = I + P + P^2 + P^3 + \ldots$? – LinAlg Oct 11 at 15:58
  • @LinAlg Yes, I've thought about this fact, but I don't see how this can be helpful... – Ludwig Oct 11 at 15:59

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.