1
$\begingroup$

Let $X=\mathbb K^n$, where $\mathbb K=\mathbb R$ or $\mathbb C$. I have seen proofs that the functions

$$\|x\|_p:=\sqrt[p]{\sum_i|x_i|^p},\qquad p\in[1,\infty]$$

are all norms. (The $p=\infty$ case must be interpreted as a limit $p\to\infty$, which turns out to be equivalent to $\max_i|x_i|$.) It is also straightforward to check that multiplying any of these functions by a positive constant results in a norm. That is,

$$\|x\|_{(p,\lambda)}:=\lambda\|x\|_p,\qquad p\in[1,\infty],\lambda>0$$

are norms. Are there any other norms?

By the way, I am aware that all norms on finite-dimensional vector spaces yield identical topologies, but this question is just about the norms.

$\endgroup$
3
$\begingroup$

Let $K$ be a compact and convex subset of $\Bbb R^n$ such that $0$ is an interior point of $K$ and that $v\in K\implies-v\in K$. Then you can define the norm $\|\cdot\|_K$ such that $\|0\|_K=0$ and that, if $v\ne0$, $\|v\|_K$ is the smallest $\lambda\in(0,\infty)$ such that $\lambda^{-1}v\in K$. With respect to this norm, $K$ is the unit ball. If, for instance $n=2$ and $K$ is an hexagon centered at the origin, $\|\cdot\|_K$ is norm which is different from any $\|\cdot\|_p$.

$\endgroup$
2
  • 1
    $\begingroup$ Your definition seems to be off: we need the smallest $\lambda \in (0,\infty)$ with $\lambda^{-1} v \in K$ (or equivalently: $v \in \lambda K$). Moreover, one should note that $K = \{v | \|v\|_K \le 1\}$ becomes the unit ball and $\|\cdot\|_K$ is called Minkowski functional of $K$. $\endgroup$
    – gerw
    May 28 at 6:54
  • $\begingroup$ @gerw I've edited my answer. Thank you. $\endgroup$ May 28 at 8:42
1
$\begingroup$

If $T:X \to X$ is any vector space isomorphism then $\|x\|=\|Tx\|_2$ (where $\|x\|_2$ is the usual norm gives you a new norm on $X$.

A specific example: $\|x\|=\sqrt {|x_1+x_2|^{2}+|x_2|^{2}+\cdots |x_n|^{2}}$.

$\endgroup$

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.