In the Wikipedia proof of the Minkowski inequality (http://en.wikipedia.org/wiki/Minkowski_inequality), the following inequality is used:
I was just wondering if this inequality has a name or if this is too "first principles" to warrant a name.


  • $\begingroup$ If I read the article correctly it's just using convexity. $\endgroup$ – Michael Hoffman Nov 2 '09 at 7:53
  • $\begingroup$ Thanks Michael, I was under the impression that this was generally true! $\endgroup$ – mornington Nov 2 '09 at 8:10
  • 1
    $\begingroup$ It is generally true. It is the composition of the triangle inequality |f+g| </= |f|+|g| composed with the power mean inequality (a+b)/2 </= ((a^p+b^p)/2)^(1/p) for a=|f| and b=|g| which is a consequence of convexity. $\endgroup$ – Philipp Lampe Nov 2 '09 at 8:19

It is not named after a person, but I would argue that the inequality does have a name, "the convexity inequality for xp". If you look at the definition of a convex function, the statement that f(x) = xp is convex is exactly the statement that ((x+y)/2)p ≤ (xp+yp)/2. (Plus either that f is continuous or that the weighted version also holds). It is also correct to say "is a consequence of convexity", but it seems better to call it the statement of convexity than a consequence of convexity.

The multivariate form of the convexity inequality is named after a person; it is Jensen's inequality.

All of this may seem like a pat answer, but it works. It is a theorem that you get a valid norm if xp is replaced by any non-negative convex function φ(x) with suitable behavior at 0 and ∞. The resulting norm is called an Orlicz norm and the resulting Banach space is called an Orlicz space.


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.