? p-norm <= (p-1)norm <= <= 2-norm <= 1-norm?

  • Thread starter Thread starter patiobarbecue
  • Start date Start date
Click For Summary

Homework Help Overview

The discussion revolves around the properties of p-norms for vectors in R^n, specifically examining the relationships between different p-norms and their validity as norms under various conditions. The original poster attempts to verify the inequality || x ||_p <= || x ||_{p-1} and questions the nature of p-norms when p is less than 1.

Discussion Character

  • Exploratory, Conceptual clarification, Mathematical reasoning, Assumption checking

Approaches and Questions Raised

  • Participants discuss the application of Jensen's inequality to establish relationships between different p-norms. Some explore specific cases, such as p = 2 and p = 3, while others question the validity of the p-norm for values of p less than 1 and whether it satisfies the properties of a norm.

Discussion Status

There is an ongoing exploration of the properties of p-norms, with some participants providing insights and potential approaches, while others raise questions about the assumptions underlying the definitions of norms. The discussion is active, with various interpretations being considered.

Contextual Notes

Participants note that the p-norm is defined for p >= 1 as a true norm, leading to questions about its status for 0 < p < 1. The implications of this distinction are under examination.

patiobarbecue
Messages
10
Reaction score
0
1.
the p-norm for a vector x in R^n is defined usually:

|| x ||_p = (x_1^p + x_2^p + ... + x_n^p)^{1/p}

the question is to verify:

|| x ||_p <= || x ||_{p-1}






Homework Equations



I guess even more generally p-norm is a decreasing function in p for "any" x?


The Attempt at a Solution



Neither Cauchy, nor the more general Holder doesn't seem to apply.
 
Physics news on Phys.org
Try using Jensen's inequality:

If \phi is a convex function, and y_1, \ldots, y_N lie in its domain, and \lambda_1, \ldots \lambda_N are positive real numbers such that

\sum_{n=1}^{N} \lambda_n = 1

then

\phi\left(\sum_{n=1}^{N} a_n y_n\right) \leq \sum_{n=1}^{N} a_n \phi(y_n)

For p = 2, try using y_n = x_n^2, \phi(x) = x^{1/2}, and a_n = 1/N. You will have to modify these choices appropriately for general p.

Also, your proposed generalization is correct: if 1 \leq q \leq p \leq \infty then ||x||_p \leq ||x||_q. The p = \infty case is handled separately (Jensen's inequality doesn't apply) but it's easy.
 
I think it's a proof for norm 2 and 3. For any n, n-1 it's the same essentially.

First consider R2.
Note (x^2+y^2)^1/2 > (x^3+y^3)^1/3 is equivalent to
[1+(y/x)^2]^1/2 > [1+(y/x)^3]^1/3
Assume y/x < 1 without losing generality, then from [1 + ...] > 1 one sees
[1+(y/x)^2]^1/2 > [1+(y/x)^3]^1/2 > [1+(y/x)^3]^1/3.

Next do an induction toward Rn.
(x1^2+x2^2+...+xn^2)^1/2 > (x1^3+x2^3+...+xn^3)^1/3 is equivalent to
(y^2+xn^2)^1/2 > (z^3+xn^3)^1/3, where y and z stands for (x1^2+...+xn-1^2)^1/2 and (x1^3+...+xn-1^3)^1/3 respectively.
By induction y > z. Hence (y^2+xn^2)^1/2 > (z^2+xn^2)^1/2 > (z^3+xn^3)^1/3. QED.
 
thanks, guys, it helps!
 
||x||_p isn't a norm for p<1 ?

Hello people! I have a question about the p-norm.
We know that the p-norm is a true norm for p >= 1.
But I dont'n understand why it is not a true norm for 0 < p < 1.
I think it hold the three conditions for a norm include when 0 < p < 1.

Somebody can helpme??
Thanks in advance!
 

Similar threads

Replies
1
Views
1K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 10 ·
Replies
10
Views
3K
Replies
9
Views
2K
  • · Replies 5 ·
Replies
5
Views
4K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 5 ·
Replies
5
Views
3K
  • · Replies 6 ·
Replies
6
Views
2K
  • · Replies 9 ·
Replies
9
Views
2K
  • · Replies 5 ·
Replies
5
Views
2K