Comparing Norms & Metrics: Axioms & Differences

In summary: For all a ∈ F and all u, v ∈ V, p(av) = |a| p(v), (positive homogeneity or positive scalability).(2) p(u + v) ≤ p(u) + p(v) (triangle inequality or subadditivity).(3) If p(v) = 0 then v is the zero vector (separates points).
  • #1
Bachelier
376
0
Are the axioms of a Norm different from those of a Metric?

For instance Wikipedia says:

a NORM is a function p: V → R s.t. V is a Vector Space, with the following properties:
For all a ∈ F and all u, v ∈ V, p(av) = |a| p(v), (positive homogeneity or positive scalability).
p(u + v) ≤ p(u) + p(v) (triangle inequality or subadditivity).
If p(v) = 0 then v is the zero vector (separates points).

While a metric is defined as a distance function between elements of a set with the 3 familiar axioms, of positive definition, triangular inequality and symmetry.

The terms are however used loosely in many math books and make you think they are one and the same.

The reason I am asking is I am faced with a question where given a norm definition on the set ##\mathbb{Q}## I am asked to determine if the operation defines a norm? So I was wondering which axioms do I check, those of the Metric or those of the VS norm?

Thank you
 
Physics news on Phys.org
  • #2
A norm and a metric are two different things. The norm is measuring the size of something, and the metric is measuring the distance between two things.

A metric can be defined on any set ##S##. It is simply a function ##d## which assigns a distance (i.e. a non-negative real number) ##d(x,y)## to any two elements ##x,y \in S##. It must satisfy three criteria in order to qualify as a metric. For all ##x,y,z\in S##, (1) ##d(x,y) \geq 0##, and ##d(x,y) = 0## if and only if ##x = y##; (2) ##d(x,y) = d(y,x)##; (3) ##d(x,z) \leq d(x,y) + d(y,z)##.

If ##S## happens to be a vector space, meaning that there is a way to add the elements together and multiply them by scalars, then we can define a norm. This is a function ##\rho## which maps each element to a nonnegative real number. It must satisfy the following rules for all ##x,y \in S## and all scalar ##\alpha##: (1) ##\rho(x) \geq 0##, and ##\rho(x) = 0## if and only if ##x = 0##; (2) ##\rho(\alpha x) = |\alpha| \rho(x)##; (3) ##\rho(x+y) \leq \rho(x) + \rho(y)##.

If we have a vector space with a norm ##\rho##, it is always possible to define a metric in terms of that norm by putting ##d(x,y) = \rho(x-y)##. But not every metric is defined in terms of a norm, or even CAN be defined in terms of a norm. An example is the metric
$$d(x,y) = \begin{cases}
1 & \textrm{ if }x \neq y \\
0 & \textrm{ if }x = y \\
\end{cases}$$
It is easy to check that this satisfies the three conditions for a metric. But even if we define this metric on a vector space, there is no norm ##\rho## satisfying ##d(x,y) = \rho(x-y)##. To see this, suppose there were such a ##\rho##, and let ##\alpha## be any scalar such that ##\alpha \neq 0## and ##|\alpha| \neq 1##. Choose any ##x,y## such that ##x \neq y##. Then ##\alpha x \neq \alpha y##, and we must have
$$1 = d(\alpha x, \alpha y) = \rho(\alpha x - \alpha y) = \rho(\alpha(x-y)) = |\alpha| \rho(x-y) = |\alpha| d(x,y) = |\alpha|$$
which is a contradiction.
 
  • #3
jbunniii said:
A norm and a metric are two different things. The norm is measuring the size of something, and the metric is measuring the distance between two things.

A metric can be defined on any set ##S##. It is simply a function ##d## which assigns a distance (i.e. a non-negative real number) ##d(x,y)## to any two elements ##x,y \in S##. It must satisfy three criteria in order to qualify as a metric. For all ##x,y,z\in S##, (1) ##d(x,y) \geq 0##, and ##d(x,y) = 0## if and only if ##x = y##; (2) ##d(x,y) = d(y,x)##; (3) ##d(x,z) \leq d(x,y) + d(y,z)##.

If ##S## happens to be a vector space, meaning that there is a way to add the elements together and multiply them by scalars, then we can define a norm. This is a function ##\rho## which maps each element to a nonnegative real number. It must satisfy the following rules for all ##x,y \in S## and all scalar ##\alpha##: (1) ##\rho(x) \geq 0##, and ##\rho(x) = 0## if and only if ##x = 0##; (2) ##\rho(\alpha x) = |\alpha| \rho(x)##; (3) ##\rho(x+y) \leq \rho(x) + \rho(y)##.

If we have a vector space with a norm ##\rho##, it is always possible to define a metric in terms of that norm by putting ##d(x,y) = \rho(x-y)##. But not every metric is defined in terms of a norm, or even CAN be defined in terms of a norm. An example is the metric
$$d(x,y) = \begin{cases}
1 & \textrm{ if }x \neq y \\
0 & \textrm{ if }x = y \\
\end{cases}$$
It is easy to check that this satisfies the three conditions for a metric. But even if we define this metric on a vector space, there is no norm ##\rho## satisfying ##d(x,y) = \rho(x-y)##. To see this, suppose there were such a ##\rho##, and let ##\alpha## be any scalar such that ##\alpha \neq 0## and ##|\alpha| \neq 1##. Choose any ##x,y## such that ##x \neq y##. Then ##\alpha x \neq \alpha y##, and we must have
$$1 = d(\alpha x, \alpha y) = \rho(\alpha x - \alpha y) = \rho(\alpha(x-y)) = |\alpha| \rho(x-y) = |\alpha| d(x,y) = |\alpha|$$
which is a contradiction.

Thank you for clarifying this: so which axioms do I have to check for the norm?
 
  • #4
Given a norm, |v|, we can define a metric: d(u, v)= |u- v|. However, there exist metric spaces which do NOT correspond to a norm. The simplest is the "discrete metric": for any set, define d(x, y)= 0, if x= y, 1, otherwise.

It is also true that, given any inner product, <u, v>, we can define a norm: [itex]|v|= \sqrt{<v, v>}[/itex], and so a metric. But there exist norms that do not correspond to any inner product.

(Edited thanks to micromass.)
 
Last edited by a moderator:
  • #5
HallsofIvy said:
But there exist inner product spaces which cannot be given norms.

I guess you mean this backwards: there exists normed spaces which cannot be given an inner product.
 
  • #6
Bachelier said:
Thank you for clarifying this: so which axioms do I have to check for the norm?
Check that it satisfies the conditions that I mentioned above:
If ##S## happens to be a vector space, meaning that there is a way to add the elements together and multiply them by scalars, then we can define a norm. This is a function ##\rho## which maps each element to a nonnegative real number. It must satisfy the following rules for all ##x,y \in S## and all scalar ##\alpha##: (1) ##\rho(x) \geq 0##, and ##\rho(x) = 0## if and only if ##x = 0##; (2) ##\rho(\alpha x) = |\alpha| \rho(x)##; (3) ##\rho(x+y) \leq \rho(x) + \rho(y)##.
 
  • #7
Norms are linear. Metrics only need to fulfil the triangle inequality. If you want to see a cool metric that cannot be derived from a norm look up the French railroad metric.
 
  • #8
If you're asked to check if a given function is a norm, you must of course (directly or indirectly) check if the function satisfies the conditions listed in the definition of "norm". Anything else would be absurd. If there had been a theorem that says that every metric is a norm, then it would of course have been sufficient to check that the given function is a metric. But there's no such theorem. The conditions that a metric is required to satisfy do not even mention a vector space structure.

Another thing worth noting is that the term "metric" is also used as a short form of "metric tensor field" in differential geometry. That kind of metric is a function that takes each point p in a manifold M to something very similar to an inner product on the tangent space of M at p. (The tangent space at p is a vector space). So I would say that inner products are more closely related to the "metrics" of differential geometry than the metrics of metric spaces.
 
  • #9
micromass said:
I guess you mean this backwards: there exists normed spaces which cannot be given an inner product.
Yes, thanks for catching that. I have gone back and edited it.
 

1. What is the difference between norms and metrics?

Norms and metrics are both mathematical concepts used to quantify and compare various characteristics or properties. However, the main difference between them is that norms are used to measure the size or magnitude of a mathematical object, while metrics are used to measure the distance or similarity between two mathematical objects.

2. How are norms and metrics related?

Norms and metrics are closely related as they both involve the use of mathematical operations to compare or evaluate different mathematical objects. In fact, metrics can be derived from norms, as most metrics are based on a normed vector space.

3. What are the axioms of norms and metrics?

The axioms of norms and metrics are a set of fundamental properties that these mathematical concepts must satisfy. Some common axioms for both norms and metrics include positivity, triangle inequality, and homogeneity.

4. What are some examples of norms and metrics?

Examples of norms include Euclidean norm, Manhattan norm, and Chebyshev norm. Examples of metrics include Euclidean distance, Manhattan distance, and Hamming distance.

5. How are norms and metrics used in real-world applications?

Norms and metrics have a wide range of applications in various fields, including mathematics, physics, engineering, and computer science. Some common uses include data analysis, machine learning, optimization problems, and image processing.

Similar threads

  • Linear and Abstract Algebra
Replies
8
Views
1K
Replies
1
Views
708
Replies
1
Views
1K
  • Differential Geometry
Replies
7
Views
2K
  • Calculus and Beyond Homework Help
Replies
19
Views
1K
  • Science and Math Textbooks
Replies
24
Views
3K
  • Topology and Analysis
Replies
2
Views
1K
  • Differential Geometry
Replies
2
Views
534
  • Calculus and Beyond Homework Help
Replies
5
Views
3K
Replies
2
Views
2K
Back
Top