Say that a function $\,f: \mathbb{R}^{n\times n} \to \mathbb{R}$ is given in an element-wise form as

$$f(A) = \sum_{i,j} A_{ij} F_{ij}.$$

When is this function convex on the whole set $\mathbb{R}^{n\times n}$?

While trying to find an answer to this question I've stumbled upon a definition of quadratic forms based on matrices, where the condition for convexity is straightforward: $f$ is convex if and only if $F$ (the matrix defining the quadratic form element-wise) is positive semi-definite. Is there a correspondence with the case of a more general matrix function like above?


$f$ is convex for all $F$ as $f(A)$ is linear in $A$.
$f(\theta A+(1-\theta)B)=\sum_{ij}(\theta A_{ij}+(1-\theta)B_{ij})F_{ij}=\theta\sum_{ij} A_{ij} F_{ij}+(1-\theta)\sum_{ij}B_{ij}F_{ij}=\theta f(A)+(1-\theta)f(B)$


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.