I need to solve the following equation for $P \in\mathbb{R}^{r\times d}$

$$P - G_1G_2(\lambda P^\top(PAP^\top)^{-1}P + A^{-1} ) = 0,$$

where the other quantities are known: $A\in\mathbb{R}^{d\times d}$, $G_1 \in\mathbb{R}^{r\times d}$, $G_{2} \in\mathbb{R}^{d\times d}$, $\lambda\in\mathbb{R}$.

I have already tried everything I know. If the equation had involved scalar quantities the solution would simply be $$ p = g_1g_2(a^{-1}(1+\lambda)), $$

but unfortunately in the matrix case I don't know how to proceed. Maybe there should be some way to decompose it using SVD or some trace trick?


1 Answer 1


The solution for $P$ to $$P - G_1G_2(\lambda P^\top(PAP^\top)^{-1}P + A^{-1} ) = 0$$ is $$P=(1 +\lambda )G_1 G_2 A^{-1},$$ as one can check by substitution into $$G_1G_2P^\top(PAP^\top)^{-1}P=G_1 G_2(G_1G_2A^{-1})^\top\bigl(G_1G_2(G_1G_2A^{-1})^\top\bigr)^{-1}G_1G_2A^{-1}=G_1G_2A^{-1}.$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.