Does the supporting hyperplane theorem imply that closure of any convex set can be expressed as intersection of halfspaces(possibly infinitely many halfspaces). The statement of supporting hyperplane theorem is: For any nonempty convex set $C$, and any $x_0 \in boundary(C)$, there exists a supporting hyperplane to $C$ at $x_0$.
Asked
Active
Viewed 1,096 times
1
-
I'm assuming you're talking about finite-dimensions? In infinite dimensions, this theorem doesn't hold (although it does hold for $x_0$ in a norm-dense subset of the boundary of $C$). It also holds under the assumption that $C$ has non-empty interior. – Theo Bendit Oct 30 '17 at 08:17
-
Thanks for the answer. I was talking about finite dimensions but would love to know a counterexample in infinite dimensional space. – Curious Oct 30 '17 at 08:27
-
Here's one: https://mathoverflow.net/questions/121526/are-all-points-x-of-the-boundary-of-a-convex-set-c-of-a-hilbert-space-h-projecti – Theo Bendit Oct 30 '17 at 08:35
1 Answers
1
By definition, the convex set $S$ is contained in one of the halfspaces bounded by its supporting hyperplane. So $S$ lies in any intersection of these halfspaces.
By the separating hyperplane theorem, any point $a$ outside of $S$ can be separated, such that the halfspace that contains $S$ does not contain $a$. So the intersection of all such subspaces leaves you with $S$.
That is, for the set of points $S^c = \{x:x \notin S\}$, there exists an intersection of halfspaces $B$ such that $S^c\cap B = \emptyset$ and $S \subseteq B$. Thus $B = S$.
Ken Wei
- 1,759