Proving Euclidean and supremum norms are equivalent on R²

  • Thread starter Thread starter gtfitzpatrick
  • Start date Start date
  • Tags Tags
    Equivalent
Join the discussion
Registration is free. Start your own thread to ask a follow-up.
11 replies · 4K views
gtfitzpatrick
Messages
372
Reaction score
0

Homework Statement



Show that the Euclidean and supremum norms are equivalent norms on R2

The Attempt at a Solution



The Euclidean Norm is
[tex]\left|\left|[/tex]X[tex]\left|\left|[/tex]1 = [tex]\sqrt{x1^2 + x2^2}[/tex]

The Supremum norm is
[tex]\left|\left|[/tex]X[tex]\left|\left|[/tex][tex]\infty[/tex] = max ([tex]\left|[/tex]x1[tex]\left|[/tex],[tex]\left|[/tex]x2[tex]\left|[/tex])

so for them to be equivalent:

a([tex]\sqrt{x1^2 + x2^2}[/tex])[tex]\leq[/tex] max ([tex]\left|[/tex]x1[tex]\left|[/tex],[tex]\left|[/tex]x2[tex]\left|[/tex]) [tex]\leq[/tex] b([tex]\sqrt{x1^2 + x2^2}[/tex])

I think I'm going right with this?
im not sure how to work with the supremum norm in the middle?
 
Physics news on Phys.org
That sounds like a correct algebraic translation of the problem. In more detail, you want to prove:



There exist positive real numbers a and b such that
For any real numbers x1 and x2
[tex]a \sqrt{x_1^2 + x_2^2} \leq \max \{ |x_1|, |x_2| \} \leq b \sqrt{x_1^2 + x_2^2}[/tex]​



The supremum in the middle is not essential -- with a little effort you can rearrange that inequality so that there are suprema is on the outside and one Euclidean norm in the middle. Really, for the purposes of solving for a and b, you probably just want to look at it as two separate inequalities.
 
Yes, what you have to prove is you can find the a and b. You might start by stating that without loss of generality, you may assume [itex]|x_1| \le |x_2|[/itex], so that:

max{|x1|,|x2|} = |x2|

and work with assumption.
 
thanks for the help lads

[tex] a \sqrt{x_1^2 + x_2^2} \leq b \sqrt{x_1^2 + x_2^2} \geq \max \{ |x_1|, |x_2| \} }[/tex]

[tex]\Rightarrow[/tex]

[tex] a \sqrt{x_1^2 + x_2^2} \leq b \sqrt{x_1^2 + x_2^2} \geq |x_2| }[/tex]

[tex]\Rightarrow[/tex]

[tex] a \leq b \geq |x_2|/\sqrt{x_1^2 + x_2^2} }[/tex]

a constant?
 
Just draw a right triangle. Make the lengths of the legs x1 and x2. sqrt(x1^2+x2^2)=h is the length of the hypotenuse. In terms of h, what's the longest the longest leg can be? What's the shortest the longest leg can be?
 
the shortest the longest leg can be is x1=x2 so then x2 = [tex]\sqrt{(h^2)/2}[/tex] ?
 
gtfitzpatrick said:
the shortest the longest leg can be is x1=x2 so then x2 = [tex]\sqrt{(h^2)/2}[/tex] ?

Exactly. h/sqrt(2). How about the longest the longest leg can be? So the max norm is bounded by constants times the euclidean norm, right?
 
the longest the longest leg can be is x1 = 0 so then x2 = h?
 
gtfitzpatrick said:
the longest the longest leg can be is x1 = 0 so then x2 = h?

I would have said 'the longest the longest can be is h'. But I think you've got the idea. If h is the hypotenuse, the longest leg is between h and h/sqrt(2).
 
ok, so the longest leg is between h and h/sqrt(2), i get that with the triangle but I am not sure how it shows the norms are equivalent?sorry and thanks for the help!
 
The hypotenuse is the euclidean norm of the point (x1,x2) in R^2. The longest leg is the max norm of (x1,x2).
 
oh god,yes-thanks a million...that actually really,really explains it to me. Damn thanks a million.Thanks dick. I am actually happy now :biggrin: