0

$L_2$ norm is defined as

$$ ||x||_2 = \sqrt{\sum_{i=1}^{n} x_{i}^2} $$

or

$$ ||x||_2 = \sqrt{x^{T}x} $$ (linear operators form)

or

$$ \langle u,v\rangle_2 = \sum_{i=1}^{N} u_{i}v_i $$ (dot product of two vectors that defines norm $L_2$ if $u=v$)

and $$ ||\phi||_{2} = \sqrt{\int\ |\phi(x)^2|dx} $$ but from definition of integal we know $$\int\ |\phi(x)^2|dx = \lim_{n\to\infty}\sum_{i=0}^{n}|\phi(x_{i})\phi(x_{i})|\delta x_{i} $$ then $$ ||\phi||_{2} = \sqrt{\lim_{n\to\infty}\sum_{i=0}^{n}|\phi(x_{i})\phi(x_{i})|\delta x_{i}}$$ that does not correspond to definition #1 if $\phi(x_{i}) = 1\cdot x_{i}$

So the confusion comes form: I do not understand where does the term $\delta x_{i} $ vanishes in the first (or third definitions compared to the last one).

In vector case we take dot product of vector with itself and it gives us its norm, however if we say that the integral above can serve as a definition of $L_2$ norm I do not understand why do not why multiply the dot product by the third vector $\delta X_{i} $. I understand that the scalar product already gives us the scalar (we do not have to multiply it by vector once again) but I do not understand how we obtain a consistency between the definition of integral and $L_2$ norm. Maybe it comes from normalization? The integral form of inner product . But then I cannot see why the boundaries for the norm are not between $0$ and $1$ all the time. Thank you!

Mårten W
  • 3,480
  • You write $|x|{2} = \sqrt{\int\ |\phi(x)^2|dx}$, but this is nonsense. I mean, what is $\phi$? Also, you have $x$ on the LHS and on the RHS $x$ is the variable of integration. Makes no sense. Instead, what's right is $|\phi|{2} = \sqrt{\int\ |\phi(x)^2|dx}$ – amsmath Oct 16 '17 at 13:27
  • @amsmath thanks! I have written that automatically. Still I do not understand why there is a definition consistency – Maksym Bondarenko Oct 16 '17 at 13:35
  • You have two completely different spaces. One is for sequences $(x_i)$ and the other for functions. Both have a norm which is called $L^2$-norm (where actually in the sequence case one coins it $\ell^2$-norm). You cannot compare these two. – amsmath Oct 16 '17 at 13:37
  • Oh I see your point. http://mathworld.wolfram.com/L2-Norm.html This link explaines "L2 norm is reserved for application with a function $\phi(x)$". So if I work in a function space the last expansion where I define the sum as limit will be true? – Maksym Bondarenko Oct 16 '17 at 13:48
  • Yes. But usually you take the Lebesgue integral, which is defined differently. – amsmath Oct 16 '17 at 15:40
  • @amsmath thanks I will look – Maksym Bondarenko Oct 17 '17 at 16:37

0 Answers0