Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Say that $V$ is a finite-dimensional real normed vector space, where for some $v \in V$ the norm is notated by $\|v\|$. Then say that $V^*$ is the dual space of linear functionals on $V$.

The "dual norm" $\| \cdot \|^*$ naturally exists on $V^*$, given for some $v^* \in V^*$ by

$\displaystyle\|v^*\|^* = \sup_{v \in V} \left\{ \frac{|\langle v^*, v\rangle|}{\|v\|} \right\} = \sup_{v \in V} \left\{ |\langle v^*, v\rangle| : \|v\| = 1 \right\}$

for $\langle v^*,v \rangle$ the dual pairing between $v^*$ and $v$.

If a basis is chosen for $V$, then is vectors can be represented as $\mathbb{R}$-tuples. Sometimes the original norm on $V$ can be denoted by a closed-form expression on the entries of this tuple. For instance, the $\ell_p$ norm on a vector $(x_1, x_2, ..., x_n)$ is given by $\left(|x_1|^p + |x_2|^p + ... + |x_n|^p\right)^{\frac{1}{p}}$.

In these cases, is there a specific algorithm for finding a closed-form expression for the dual norm given the dual basis of $V^*$?

(Obviously the solution is well known for the $\ell_p$ norm, but what about in general?)

share|improve this question

1 Answer 1

up vote 1 down vote accepted
+50

There is no algorithm to find a closed form expression for the dual norm, even in two dimensions. For example, let the unit disk be $\{(x,y): x^2+|y|+p(x^2) \le 1 \}$ where $p$ is a polynomial of high degree with sufficiently small coefficients so that the unit disk is convex. Finding the dual norm of vector $(t, 1)$ ($t$ small) leads to the search for point on the graph of $y=1-x^2-p(x^2)$ with slope $-t$. This amounts* to solving $2x+2xp'(x^2)=t$. Since $p$ can be a generic polynomial of high degree (as long as the coefficients are small), there is no closed form for the solution.

(*) One can imagine finding the minimum without equating the derivative to zero; but the methods for doing so are algorithmic and do not produce a closed-form expression for the minimum in terms of $t$.

share|improve this answer
    
This is a great point. Thanks for this. Do you know of any conditions on the norm which are sufficient to deem the dual norm to have a closed form solution? I'm stuck specifically trying to figure out the various dual norms of norms which can be obtained by taking quotient and subspace norms of $\ell_p$ norms (say that five times fast...) –  Mike Battaglia Oct 8 '13 at 21:31

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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