Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $S\subset \mathbb{R}^n$ and let $f(x)$ be a continuous function over $\mathbb{R}^n$. Furthermore, define $s_{\text{max}}:= \sup_{x\in S} \{f(x)\}$ and let $f(x)$ attain its minimum for at least one element in $S$.

Under what conditions on $f(x)$ and $S$ does it hold that $f(x_1)\leq s_{\text{max}}$ $\Rightarrow$ $x_1\in S$?

In particular, does this hold if $f(x)$ is a convex function and $S$ is a convex set? If $f(x)$ is convex and $S$ is an arbitrary set?

Unfortunately I am far from my comfort zone here, so any help is much appreciated!

share|cite|improve this question
Do you mean $f(x) < s_{max} \implies x \in S$ or $f(x) < s_{max} \Leftrightarrow x \in S$? In the latter case, no such function exists as then $x_{max} \notin S$ and in the former, for any set any constant function works. You might have wanted $\leq$ instead. Also, some sets don't have a $\max$, you might want to use $\sup$ instead. – Karolis Juodelė Aug 28 '12 at 17:27
I am only interested in $f(x)\leq s_{max} \Rightarrow x\in S$. Thanks for your comments, I have amended the question based on them. Yes, a constant function would work but I would like to know what other possibilities are allowed. – Juan Miguel Arrazola Aug 28 '12 at 18:43
up vote 1 down vote accepted

Let $\Sigma_S = \{x | f(x) \leq \sup_{t \in S} f(t) \}$. If $x \in S$, then we have $f(x) \leq \sup_{t \in S} f(t)$, hence $x \in \Sigma_S$, i.e, $S \subset \Sigma_S$. This is true for all $S,f$.

The desired property above is equivalent to $\Sigma_S \subset S$, hence this is true iff $S = \Sigma_S$.

That is, the property you desire is true iff $S$ has the form $\{x | f(x) \leq L \}$ for some $L$ (which may be $\infty$).

For a counterexample, choose the convex function $f(x) = x^2$ on the convex set $[0,1]$. It is easy to see that $s_\max =1$, but $f(-1) \leq 1$ also.

share|cite|improve this answer
This is precisely what I was looking for. Many thanks! – Juan Miguel Arrazola Aug 28 '12 at 22:32

I'll try answering the following question. What kind of set $S \subset \mathbb{R}^n$ can be defined as $\{x|f(x) \leq c\}$ for a continuous $f : \mathbb{R}^n \rightarrow \mathbb{R}$ and constant $c$.

Trivially, the function $f(x) = \inf_{y \in S} d(x, y)$ - distance from $S$ to $x$ defines any closed set $S$. In general the shape of $f$ can vary a lot as long as $f(\partial S) = c$ and $f$ has the right relation to $c$ in the interior and exterior of $S$. Open sets would work if the condition was $f(x) < c$. Now $f$ would measure the (negative of) distance from $x$ to the complement of $S$ instead.

This does not cover all sets though. $\mathbb{Q} \subset \mathbb{R}$, for example is neither closed nor open. It is clear that unless $f(\mathbb{Q}) = c$ the function will not be continuous. Now take any Cauchy sequence $(x_n)$ that does not converge in $\mathbb{Q}$. Due to continuity of $f$ we have that $f(\lim x_n) = \lim f(x_n) = \lim c = c$ - a contradiction.

share|cite|improve this answer

Let's assume that $S$ is a subset of the domain of $f$ which -- for the case of maximisation -- we define as $\operatorname{dom}f=\{x|f(x)>-\infty\}$. Then, the fact that $f(x)\leq s_{max}$ does not imply that $x\in S$ nor that $x\notin S$. You can see this through a simple example. Consider the concave function $f(x)=-x^2$ and the convex set $S=[-1,1]$. Then $s_{max}=\sup_{-1\leq x \leq 1}-x^2=0$ and take $x_0=2\notin S$. Notice that $f(x_0)\leq s_{max}$. You can easily find counter-examples for your statement using convex functions. Take for instance the convex function $f(x)=x^2$ and $S=[1,2]$ and $x_0=0$.

One useful things about convex minimisation over convex sets, is that any local minimum you find is also global. The same holds true for maximisation problems when the function is concave and the set convex.

I can only think of an extremely special case where your statement holds true. $f:\mathbb{R}^n\to\mathbb{R}$ is quadratic, i.e. $f(x)=x'Qx$ for some positive definite matrix $Q$, and $S$ is of the form $S=\{x\in\mathbb{R}^n|x'Px\leq \gamma\}$ for some $\gamma>0$ and $P=P'$ positive definite.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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