Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I've been trying to prove the concavity of a particular function which I reduced to proving the reverse Minkowski Inequality for $p \le 1$, $p \ne 0$ for arguments in $\mathbb{R}^{n}_{+}$. That is,

$$ \|x \|_p + \|y\|_p \le \|x+y\|_p $$

I've found a proof for the case of $0 < p \le 1$, but the case when $p < 0$ still evades me. All proofs of Minkowski's Inequality (in the proper direction) usually rely on Hölder's Inequality, which in turn relies on Young's Inequality. However, Young's does not apply for exponents below $0$, and I am rather jammed up finding another way. Can anyone offer a little direction?

share|cite|improve this question
Isn't that false, at the very least for $p=1$? Together with the usual Minkowski inequality you deduce that $$ \| x+y \| _p = |\ x\|_p + \| y \|_p ,$$ which is false generally. – Ragib Zaman Oct 17 '11 at 9:12
Quite right you are if either $x,y < 0$. I've amended it to only consider values with elements greater than or equal to 0. – duckworthd Oct 17 '11 at 21:28

Hölder is reversed when $p<0$. The proof is a simple change of variables. See Bullen's book "Handbook of Means and Their Inequalities" p.202-3 for details. Hence Minkowski is also reversed.

share|cite|improve this answer

Using the concavity of the function $f$ defined by $f(x)=x^p$, $0<p<1$, we get that for $t\in (0,1)$, $$(x_k+y_k)^p=\bigg(t\dfrac{x_k}{t}+(1-t)\dfrac{y_k}{1-t}\bigg)^p\geq t\dfrac{x_k^p}{t^p}+(1-t)\dfrac{y_k^p}{(1-t)^p}.$$ Taking the sum, we get that $$||x+y||_p^p\geq t\dfrac{||x||_p^p}{t^p}+(1-t)\dfrac{||y||_p^p}{(1-t)^p}.$$ By choosing $t=\dfrac{||x||_p}{||x||_p+||y||_p}$ and $1-t=\dfrac{||y||_p}{||x||_p+||y||_p}$, we get that

\begin{eqnarray*} ||x+y||_p^p &\geq& t\dfrac{||x||_p^p}{\dfrac{||x||_p^p}{\bigg(||x||_p+||y||_p\bigg)^p}}+(1-t)\dfrac{||y||_p^p}{\dfrac{||y||_p^p}{\bigg(||x||_p+||y||_p\bigg)^p}}\\ \Rightarrow ||x+y||_p^p &\geq& t\bigg(||x||_p+||y||_p\bigg)^p+(1-t)\bigg(||x||_p+||y||_p\bigg)^p=\bigg(||x||_p+||y||_p\bigg)^p\\ \Rightarrow||x+y||_p &\geq& ||x||_p+||y||_p. \end{eqnarray*}


share|cite|improve this answer

Your Answer


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.