14
$\begingroup$

I am confused about the difference between coordinates and basis. My confusion is following:

Let $e_i$ denote the standard basis and $v_i$ denote a non-standard basis of a finite $n$-dimensional vector space $V$. Then $e_i = (\delta_{ij})$ (all entries zero except $i$-th). But: the coordinates of $v_i$ with respect to basis $v_i$ also are $(\delta_{ij})$. Now the definition of standard basis becomes circular: it is supposed to be the basis with vectors $(\delta_{ij})$ but every basis vector has these coordinates with respect to itself.

So what exactly is standard basis?

$\endgroup$
3
  • 2
    $\begingroup$ The coordinates of $v_i$ w.r.t. $e_i$ are not necesserily equal to $\delta_{ij}$. Maybe w.r.t. $v_i$? $\endgroup$ Aug 14, 2013 at 8:31
  • $\begingroup$ @MarkusMayr Yes! You are right! I corrected it! Thank you! $\endgroup$
    – blue
    Aug 14, 2013 at 8:44
  • 1
    $\begingroup$ Vector spaces admit many bases in general. But there are many specific families of vector spaces that come equipped with one $-$ a "standard" one. E.g. $K^n$ or $K[x]$. $\endgroup$
    – anon
    Aug 14, 2013 at 8:44

5 Answers 5

13
$\begingroup$

As other answers have already pointed out, the notion of "standard basis" only applies to the very particular $K$-vector spaces $K^n$, whose elements are $n$-tuples of elements of$~K$ (scalars). I've emphasised "are", because in any $n$-dimension $K$-vector space you can represent vectors by $n$-tuples of scalars (after having chosen an ordered basis; the scalars are the coordinates of these vectors in this basis); however in general a vector and its $n$-tuple of coordinates remain two different things.

You know what the standard basis of $K^n$ is (and it is actually an ordered basis: more than just a set of vectors, it is a list where each basis vector has its own place). The one point I would like to add is mention the property that makes this particular basis stand out among other bases, either of $K^n$ or of other spaces.

Property. Any $v\in K^n$ is equal to the $n$-tuple of coordinates of $v$ with respect to the standard basis of$~K^n$.

Clearly for such a property to hold, it is necessary that such vectors $v$ be $n$-tuples of scalars in the first place, in other words that the vector space in question be $K^n$. Moreover, the coordinates $(c_1,\ldots,c_n)$ of $v$ with respect to an ordered basis $(\mathbf e_1,\ldots,\mathbf e_n)$ are by definition the scalars such that $v=c_1\mathbf e_1+\cdots+c_n\mathbf e_n$ holds; if (with $v\in K^n$) one requires that moreover $v=(c_1,\ldots,c_n)$ as stated in the property, this requires that $$ (c_1,\ldots,c_n)=c_1\mathbf e_1+\cdots+c_n\mathbf e_n. $$ This should hold for all possible values of $c_1,\ldots,c_n\in K$. In particular one can take some $c_i=1$ and all other $c_j$ zero, and this leads to the conclusion that $\mathbf e_i\in K^n$ has to be $(0,\ldots,0,1,0,\ldots,0)$ with the nonzero component at position$~i$; this should be so for every$~i$. Once these cases are checked, all others follow by linearity. So the stated property holds for the standard basis, and also characterises the standard basis of$~K^n$.

$\endgroup$
6
  • 2
    $\begingroup$ Why did my lecturer not take the time to explain this? Thanks for this. It is actually quite subtle. $\endgroup$
    – user85798
    Nov 20, 2015 at 3:43
  • $\begingroup$ @user85798 It is indeed very subtle, and while a good starting point, this answer doesn't cover everything $\endgroup$ Oct 27, 2021 at 13:49
  • $\begingroup$ @étale-cohomology Thanks for the praise. I'm looking forward to your answer covering the points I missed. $\endgroup$ Oct 31, 2021 at 5:09
  • $\begingroup$ I didn't meat to irritate you, I'm sorry, I learnt a good deal from your answer and I can't improve on it $\endgroup$ Oct 31, 2021 at 10:08
  • 1
    $\begingroup$ @GaurangAgrawal: Yes $K$ is supposed to be a field when one is talking about $K$-vector spaces. But there is no such thing as "the coordinates" (presumably living in $K$) in a general $K$-vector space (instead the elements of $K$ are used when forming linear combinations). And $K^n$ is the space whose elements are $n$-tuples of elements of $K$ (i.e., of scalars), and as my answer says it it this (and only this) $K$-vector space that has a standard basis. What is more general (than you might think) is not the notion of scalar, but the notion of vector space. $\endgroup$ Jan 8 at 15:32
9
$\begingroup$

The term standard basis only applies to vector spaces of the form $\Bbb F^n$, when every vector is of the form $(x_1, x_2, ..., x_n)^T$. You then stipulate $e_i := (0, ..., 0, 1, 0, ...,0)^T$ ($1$ in the i$^{th}$ place), which is independent of a choice of basis. For arbitrary vector spaces it doesn't really make sense to talk about a standard bases.

Edit: If $V$ is an $n$-dimensional vector space, then the coordinate vector $[v]_B$ of $v \in V$ w.r.t. a basis $B = \{v_1, ..., v_n\} $ of $V$ is defined as follows: Let $f: B \to \Bbb F^n$ be the map $v_i \mapsto e_i$. Extending $f$ to a linear map $V \to \Bbb F^n$, we define $[v]_B := f(v)$. Of course, by this definition $[v_i]_B = e_i$, but that does not mean $v_i = e_i$.

$\endgroup$
2
  • $\begingroup$ I thought I understood but then I thought about it again and now I don't: Isn't $(0,...0,1,0,...0)$ the coordinates of a basis vector $b_i$ with respect to the basis $\{b_i\}_i$? Maybe the meaning of stipulate is not too clear to me. $\endgroup$
    – blue
    Aug 14, 2013 at 12:42
  • $\begingroup$ @blue: $(0, \ldots, 0, 1, 0, \ldots, 0)$ is the coordinate vector of $b_i$ in the basis $\{b_i\}$, you're right about that. But for the standard basis $\{e_i\}$ of $\mathbb{F}^n$, it is not only the coordinate vector of $e_i$ in the basis $\{e_i\}$ that is equal to $(0, \ldots, 0, 1, 0, \ldots, 0)$, even $e_i$ itself is equal to $(0, \ldots, 0, 1, 0, \ldots, 0)$. $\endgroup$ Aug 14, 2013 at 15:25
7
$\begingroup$

"A finite $n$-dimensional vector space $V$" doesn't have a standard basis. The specific $n$-dimensional vector space $\mathbb{R}^n$ does have a standard basis: the one whose vectors are the $n$-tuples that have zero entries except for a single one.

$\endgroup$
7
$\begingroup$

There are two senses of "coordinates". One is defined with respect to a basis on a general $n$-dimensional vector space. The other refers to ordered tuplets of numbers $(a_1,\ldots,a_n)\in \mathbb R^n$. (Or whatever your base field is.) The standard basis is the unique basis on $\mathbb R^n$ for which these two kinds of coordinates are the same.

Edit: Other concrete vector spaces, such as the space of polynomials with degree $\leq n$, can also have a basis that is so canonical that it's called the standard basis. In this case, the standard basis vectors are $1, x, x^2, \ldots, x^n$.

I'm not sure that it's common to speak of a "standard basis" on a completely abstract $n$-dimensional vector space. If you have a textbook that does so, it would help to quote the context.

$\endgroup$
6
  • 2
    $\begingroup$ Just a note: sometimes the basis ${1,x,x^2,...,x^n}$ of the space of polynomials with degree $\leq n$ is called the "standard" basis. $\endgroup$
    – pppqqq
    Aug 14, 2013 at 8:37
  • $\begingroup$ @pppqqq Excellent point, I'll edit my answer. $\endgroup$ Aug 14, 2013 at 8:37
  • $\begingroup$ I understand now I think! Standard basis only has meaning in a concrete example of vector space! Thank you!! $\endgroup$
    – blue
    Aug 14, 2013 at 8:42
  • 1
    $\begingroup$ @blue: To emphasize, "most" concrete vector spaces do not have a "standard basis". Think of a line through the origin in $\mathbf{R}^2$, for example. As others note, the term "standard basis" really refers to very specific spaces, particularly $\mathbf{R}^n$ with its standard vector space operations. $\endgroup$ Aug 14, 2013 at 10:14
  • $\begingroup$ @AndrewD.Hwang After thinking about it some more I realize I don't understand after all. The expression $(0,0,....,0,1,0,...0)$ seems to be coordinates. In particular $(1,0,0,....)$ is the coordinates of a basis vector $b_1$ with respect to basis $\{b_i\}_i$. But what are standard basis vectors then? $\endgroup$
    – blue
    Aug 14, 2013 at 12:45
3
$\begingroup$

The standard basis arises when we identify a finite-dimensioned vector space $V$ with $\Bbb R^n$. We take any basis in $V$, say, $\vec v_1,\dots,\vec v_n$. Then we can say that any vector $\vec w\in V$ is a linear combination of basis vectors: $$\vec w=\sum_{j=1}^n\alpha_{j}\vec v_j.$$ In other words, instead of saying "a vector $\vec w$" we can say "a vector whose coordinates are $(\alpha_1,\dots,\alpha_n)$", which is clearly can be seen as an element in $\Bbb R^n$. Then, in $\mathbb R^n$ we introduce a standard basis.

$\endgroup$
1
  • $\begingroup$ I am sorry, it was a typo in the question. I wanted to write with respect to $v_i$. $\endgroup$
    – blue
    Aug 14, 2013 at 8:43

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.