5
$\begingroup$

Let us consider $C[0,1]$, the space of continuous functions $f\colon [0,1] \to \mathbb{R}$. It comes usually with the metric of the maximum, or of the supremum, $d_{L^{\infty}}$. Each element $f$ in $C[0,1]$ corresponds uniquely to a compact subset of $\mathbb{R}^2$, by taking the graph of $f$. Thus, we can consider the Hausdorff metric $\rho$ induced on $C[0,1]$.

My question is: "Is the Hausdorff metric $\rho$ on $C[0,1]$ equivalent to the (usual) $L^{\infty}$-metric?". (equivalent metrics for me means the underlying topologies are the same).

I understand that $\rho \leq d_{L^{\infty}}$. Also, the two metrics are not uniformly equivalent, indeed there exists no universal constant $C$ such that $d_{L^{\infty}} \leq C \cdot \rho$. To see this, just take the sequence of functions $f_n(x)= nx$ and $g_n(x)= nx+1$. The infinity distance of $f_n$ to $g_n$ is constantly equal to $1$, but the Hausdorff distance equals $1/n$.

(I am sorry if the question is too trivial/elementary, feel free to cancel it).

$\endgroup$

1 Answer 1

2
$\begingroup$

Yes, this is true.

You need to show that for any $\varepsilon>0$ and any function $f$ in $C[0,1]$ there is $\delta>0$ such that $\varepsilon$-neighborhood in one metric contains $\delta$-neighborhood in the other metric and the other way around.

One way you know already; it follows since $\rho\le d_{L^\infty}$.

The other way it follows since any continuous function $[0,1]$ is uniformly continuous. Indeed, it is easy to see that if $\omega$ is a modulus of continuity of $f$ and $\rho(f,g)<\varepsilon$ then $d_{L^\infty}(f,g)<\omega(\varepsilon)+\varepsilon$.

$\endgroup$
3
  • $\begingroup$ Thanks for your answer. Can you elaborate? I do not see your point. $\endgroup$
    – calc
    Nov 21, 2013 at 19:39
  • 1
    $\begingroup$ Ok, I think I see the point. For any $\epsilon$ infinity-ball $B$ around $f$, you want a $\delta$ $\rho$ ball around $f$ that is inside of B. You take as $\delta$ the $\delta_1/2$, where $\delta_1$ is the output of uniform continuity of $f$ for the input $\epsilon$. (for simplicity, here I am using the $1$-metric on $\mathbb{R}^2$, otherwise the formula for $\delta$ in terms of $\delta_1$ would be more complicated.) $\endgroup$
    – calc
    Nov 21, 2013 at 20:51
  • $\begingroup$ Thanks also for the revision, I think it is more clear now. $\endgroup$
    – calc
    Nov 22, 2013 at 16:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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