Let $(X,d)$ be a separable complete geodesic metric space and let $K$ be a compact (non-empty) subset of $X$. Without assuming things like linearity, the convexity of $K$, and locally convexity, under what conditions can we guarantee that there is a continuous selection for the metric projection problem: $$ \operatorname{argmin}_{k \in K} d^2(x,k)? $$

  • 1
    $\begingroup$ This rules out many natural conditions -- will you give an example of a condition of the form that you're looking for? $\endgroup$
    – Matt F.
    Jan 23 '21 at 14:11
  • $\begingroup$ Unfortunately that's the trouble, I'm only familiar with the conditions I ruled out or those for Micheal's Selection theorem? $\endgroup$ Jan 23 '21 at 14:16

In general, $argmin$ is not continuous. Even on the real line, if I take $K$ to be two distinct points, say $K=\{-1,1\}$, then $argmin_{k\in K} d^2(x,k)$ is not continuous. This is why convexity is so important. Without assuming convexity, little is guaranteed in this case.


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.