1
$\begingroup$

I need to know which permutation-invariant norms can be consistently decomposed in the sense that for any vector $v = (a,b,c)$ we have that $$\|(a,b,c)\| = \|(\|(a,b)\|,c)\|.$$

More precisely, let $v = \sum_{i=1}^n v_ie_i$ be a finite-dimensional vector, and $\{P_j\}_{j=1}^k$ a partition of the index set $\{i\}_{i=1}^n$ into $k$ subsets, so that $v = \sum_{j=1}^k \sum_{i \in P_j} v_ie_i$. The question is then for which norms is it true that for all vectors and partitions $$ \|v\| = \left\|\sum_{j=1}^k \Bigg\|\sum_{i \in P_j} v_ie_i\Bigg\|e_j\right\|. $$

It is easy to see that this is true for every $p$-norm, and every other norm that I've tried failed to have this property, so it would be natural to conjecture that $p$-norms are the only consistently decomposable ones. Just finding a counterexample to this conjecture would be very useful.

$\endgroup$
10
  • 1
    $\begingroup$ With your actual definition, there are certainly other examples such as $\|(a,b,c)\| = |a|+|c| + 2|b|$, $P_1=\{ 1, 3\}$, $P_2 = \{ 2\}$. But maybe it doesn't really formalize what you actually wanted. You could also ask for this property for all partitions, and the $p$ norms satisfy that too, of course. $\endgroup$ Commented Feb 6, 2018 at 20:36
  • 1
    $\begingroup$ Google perfectly homogeneous bases. $\endgroup$ Commented Feb 6, 2018 at 22:16
  • 1
    $\begingroup$ Your reformulation in terms of basis vectors suggests that your equation $\lVert(a, b, c)\rVert = \lVert(\lVert(a, b)\rVert, c)\rVert$, which doesn't seem to make sense for $\lVert\cdot\rVert$ interpreted as a norm on a space of fixed dimension (since the left-hand side as written takes norms of vectors in $k^3$, and the right-hand side of vectors in $k^2$), is actually supposed to be $$\lVert(a, b, c)\rVert = \bigl\lVert(\lVert(a, b, 0)\rVert, \lVert(0, 0, c)\rVert, 0)\bigr\rVert$$ (with $n = 3$, $P_1 = \{1, 2\}$ and $P_2 = \{3\}$). Is that correct? $\endgroup$
    – LSpice
    Commented Feb 7, 2018 at 0:17
  • 1
    $\begingroup$ Or just look at Theorem 1.b.7 in Lindenstrauss-Tzafriri, Classical Banach spaces, volume II $\endgroup$ Commented Feb 7, 2018 at 5:35
  • 1
    $\begingroup$ Thanks a lot, Mikhail Ostrovskii, this is exactly what I needed. I'll write up your comment as a proper answer later. $\endgroup$ Commented Feb 7, 2018 at 9:05

1 Answer 1

0
$\begingroup$

The theorem by Bohnenblust is not exactly what I wanted, but it's what I need. Adapting its statement, we have

Let $\|\cdot\|:\mathbb R^N \to \mathbb R$ be a permutation-invariant norm for $N \ge 3$ such that

  • $\|x+y\| = \|(\|x\|,\|y\|)\|$ for all $x,y$ with disjoint support.
  • $\|(1,1)\| \neq 1 $.

Then for any rationals $|a|$ and $|b|$ we have that $\|(a,b)\| = (|a|^p + |b|^p)^{\frac1p}$ for some real number $p \ge 1$.

Proof: Let $f$ be such that $f(1)=1$ and $f(n+1) = \|(1,f(n))\|$. We want to show that $f(n+m) = \|(f(m),f(n)\|$. Assume that it holds for some $m$. Then $$f(n+1+m) = \|(f(m),f(n+1)\| = \|(f(m),1,f(n)\| = \|(f(m+1),f(n))\|.$$ Since it holds for $m=1$ by definition, by induction it holds for all $m$. From that it is easy to see that $f(nm) = f(n)f(m)$ and therefore $f(n^k) = f^k(n)$. We also need to show that $f(n)$ is monotonous. This follows from applying the triangle inequality to the identity $$2(f(n),0) = (f(n),1) + (f(n),-1),$$ which implies that $f(n) \le f(n+1)$.

Now let $m,n\ge 2$ be some fixed integers, and $h$ the integer such that for any positive integer $k$ $$m^h \le n^k < m^{h+1}.$$ Using the properties of $f(n)$, it follows that $$h\log f(m) \le k \log f(n) < (h+1) \log f(m),$$ and elementary manipulations with $h$ and $k$ let us conclude that $$\frac{\log f(m)}{\log m} = \frac{\log f(n)}{\log n},$$ which means that this fraction is a constant independent of $n$ and different than $0$. Calling this constant $1/p$, we conclude that $$f(n) = n^\frac1p.$$ Now for any positive rational $m/n$ we have that $$\|(1,m/n)\| = \frac1n\|(n,m)\| = \frac1n\|(f(n^p),f(m^p)\| = \frac1nf(n^p+m^p) = (1+(m/n)^p)^\frac1p,$$ so by homogeneity $\|(a,b)\| = (|a|^p + |b|^p)^{\frac1p}$ for any rationals $|a|$ and $|b|$. I guess I can't extend this for all reals without some continuity condition.

$\endgroup$
7
  • $\begingroup$ Where are your norms defined? In an arbitrary normed space, permutation-invariance of the norm, or disjointness of the supports of vectors, just don't make sense. And perhaps you are considering some family of norms on different spaces without saying so? $\endgroup$ Commented Feb 24, 2018 at 12:52
  • $\begingroup$ Indeed, it doesn't make any sense, so that's why I'm not talking about an arbitrary normed space. I thought the question already made clear that I'm talking about a fixed norm in $\mathbb R^n$. $\endgroup$ Commented Feb 25, 2018 at 17:43
  • $\begingroup$ But then you cannot, as in your answer, insert $1^{(n)}$ for different $n$. $\endgroup$ Commented Feb 25, 2018 at 17:58
  • $\begingroup$ Sure I can. I'm using the convention that e.g. $\|(a,b)\| := \|(a,b,0)\|$. Just pad the vectors with zeroes. $\endgroup$ Commented Feb 25, 2018 at 19:47
  • $\begingroup$ Then we must have $n=3$, and you can't insert $1^{(4)}$. $\endgroup$ Commented Feb 25, 2018 at 20:56

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .