0
$\begingroup$

There is a theorem stating that every metric space embeds isometrically into $\ell _{\infty}$.

My question: is there a generalized result for pseudo metric spaces embedding isometrically into semi-normed spaces? Is/would this result be interesting/useful?

Thank you very much for taking the time to answer my question.

$\endgroup$
7
  • $\begingroup$ It is definitely not the case that every metric space embeds into $\ell^\infty$ by cardinality reasons, there must be some size assumption on the spaces $\endgroup$ Commented Oct 16, 2024 at 16:44
  • $\begingroup$ @AlessandroCodenotti certainly OP means $\ell^\infty$ of some set (possibly not countable). $\endgroup$ Commented Oct 16, 2024 at 16:48
  • $\begingroup$ Would you say what is meant by "pseudo"? Allows zero distance? negative distance? If it just relaxes to allow distance zero, I guess one can get results as trivial corollaries of the metric case. $\endgroup$ Commented Oct 16, 2024 at 16:50
  • 2
    $\begingroup$ @AlessandroCodenotti I’m pretty sure YCor meant that the index set $Y$ of $\ell_\infty(Y)$ can depend on $X$, so the second one. $\endgroup$ Commented Oct 16, 2024 at 16:56
  • 2
    $\begingroup$ @DavidGao Yes, exactly. Fix $x_0\in X$, and $Y$ a dense subset of $X$, and take the map $x\mapsto f_x\in\ell^\infty(Y)$, where $f_x(y)=d(y,x)-d(y,x_0)$. $\endgroup$ Commented Oct 16, 2024 at 17:03

1 Answer 1

4
$\begingroup$

Yes. If $(X, d)$ is a pseudometric space, then we may may define an equivalence relation $\sim$ on $X$ by $x \sim y$ iff $d(x, y) = 0$. Then $(X/{\sim}, d’)$ with $d’([x], [y]) = d(x, y)$ is a metric space, so there is some isometric embedding $\phi: X/{\sim} \to \ell_\infty(I)$. Now, let $\kappa = \sup_{x \in X} |[x]|$ and pick a vector space $V$ of cardinality at least $\kappa$ (say, $V = \mathbb{K}[\kappa]$). For each equivalence class $[x]$, fix an injection $\psi_{[x]}: [x] \to V$. Now, equip $\ell_\infty(I) \times V$ with the seminorm:

$$p(f, v) = \|f\|_{\ell_\infty(I)}$$

Then $\pi: X \to (\ell_\infty(I) \times V, p)$ defined by,

$$\pi(x) = (\phi([x]), \psi_{[x]}(x))$$

is an isometric embedding. (I don’t see how such a result is in any way interesting or useful though.)

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.