Is there a convex function $F$ that is not differentiable, but whose subdifferential admits a continuous selection, i.e. a continuous $g$ with $g(x) \in \partial F(x)$ for all $x$ in the domain?

In one dimension I think I can prove there is not: if $|\partial F(x)| > 1$ then it contains an open set, whose inverse image under $g$ is an open set, so $g$ is not a selection after all - I think it maps some $y \neq x$ to a point in the interior of $\partial F(x)$, which cannot be a subgradient at $y$.


The answer is no.

See Rockafellar's Convex Analysis, part V.

First, let $D$ be the set of points where $F$ is differentiable. Theorem 25.5 proves that $D$ is dense in the interior of the domain of $F$, with measure zero complement. And that $\nabla F$ is a continuous mapping on $D$.

Next, if the domain of $F$ has non-empty interior, then at every interior point the subdifferential has the property (Theorem 25.6):

$$ \partial F(x) = \mathrm{cl} ( \mathrm{conv}(S(x))) $$

where $S(x)$ is the set of limits of all sequences of the form $\nabla F(x_i)$ where $x_i \to x$ and $x_i \in D$.

Now if $\partial F$ admits a continuous section $g$, then for every $x$ in the interior of the domain of $F$, we have that for every sequence $x_i \in D$ such that $x_i \to x$ it holds that $g(x_i) \to g(x)$ (continuity of $g$); but since $\partial F(x_i) = g(x_i)$ is unique, this implies that $S(x) = \{g(x)\}$.

And hence $F$ is differentiable at $x$.


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.