3
$\begingroup$

A well known elementary formulation of Holder's Inequality can be stated as follows:

Let $a_{ij}$ for $i = 1, 2, \dots, k; j = 1, 2, \dots, n$ be positive real numbers, and let $p_1, p_2, \dots, p_k$ be positive real numbers such that $\sum_{i=1}^{k} \frac{1}{p_i} = 1$. Then we have

$$\sum_{j = 1}^{n} \prod_{i = 1}^{k} a_{ij} \leq \prod_{i=1}^{k} \left ( \sum_{j = 1}^{n} a_{ij}^{p_i} \right )^{\frac{1}{p_i}}.$$

Does anyone know when equality occurs in this inequality? Any insights on the equality case and/or a proof would be appreciated.


I have seen several proofs of the two sequence case. Here is one, from Cvetkovski's Inequalities: enter image description here

Perhaps this argument can be generalized to prove the above result. Thanks in advance!

$\endgroup$
1
$\begingroup$

Since the usual proof of generalized Holder's proceeds by induction from the usual Holder's inequality, the equality case is precisely the same. In other words, if $v_j$ denotes the vector with components $a_{ij}^p$, then equality occurs iff all the vectors $v_j$ are parallel.

$\endgroup$
  • $\begingroup$ Thank you -- could you direct me to a source where I can see the details of the proof? $\endgroup$ – WeierstrassSauce Jan 3 '15 at 19:55
  • 1
    $\begingroup$ I think most first-year real analysis courses cover a proof of the generalized integral version. It was an exercise in our book (Folland's Real Analysis). Someone wrote up a solution to that exercise here: math.rochester.edu/people/grads/clungstr/real_analysis/… $\endgroup$ – pre-kidney Jan 3 '15 at 20:00

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.