Okay, me previous question Can a vector space be reconstructed form its norm? was solved by a nice (but in hindsight embarassingly simple) counter-example. Perhaps a better question emerges from that asnwer as well as from the discussions in the comments there:

Let $(V,+,\cdot,\|\cdot\|)$ be a normed $\Bbb R$ vector space. Assume we are only given the underlying set $V$, the zero vector $0$, and the induced metric $d\colon V\times V\to\Bbb R$, $(v,w)\mapsto\|v-w\|$. Can we reconstruct $+$, $\cdot$, and $\|\cdot\|$ from this?

The last part is of course trivial because $\|v\|=d(v,0)$. On the other hand, if we were not given $0$ as a base point, we'd be doomed because $d$ is translation invariant. (Or we might instead be satisfied with reconstructing the affine structure)

At least in case of the Euclidean norm on a finite-dimensional space, the answer is Yes: First, we can determine $\dim V$ as one less than the maximal number of vertices in a regular simplex. If we pick such a maximal simplex with side length $1$, say, and with one vertex $0$, then we can "triangulate" all other points of $V$ by their distances to the vertices, and this allows us to reconstruct $+$ and $\cdot$.

Remains the question, what the situation looks like in all other cases. Actually, this single question automatically splits into two parts:

Q1: What is the situation with other norms on finite-dimensional spaces? "Triangulation" becomes a non-trivial task and even finding a "nice" simplex may prove difficult.

Q2: What is the situation with infinite dimensional spaces? Those can be even weirder ...

When pressed to pick one, I'd be more interested in answers to Q1 ...

EDIT: To clarify, added $\Bbb R$ as scalar field.

  • 1
    $\begingroup$ How is one expected to know the scalar field (with respect to which $V$ is a vector space)? $\endgroup$
    – hardmath
    May 4 '20 at 21:29
  • $\begingroup$ Operation $+$ is given a priory, without it there's nothing to do, how do you take $v-w$? $\endgroup$
    – janmarqz
    May 4 '20 at 21:33
  • $\begingroup$ @janmarqz I do $v-w$ to compute $d$ from $\|\cdot\|$. But if I forget the other structure and remember only $d$, I cannot immediatly subtracr $\endgroup$ May 4 '20 at 22:00

I'll assume the scalar field is $\mathbb R$. The Mazur-Ulam theorem asserts that any surjective isometry of normed linear spaces is affine. So (up to choice of origin) the metric does uniquely determine the vector space structure.


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.