Let $V$ be a finite dimensional vector space and let $B=\{v_1,v_2,\cdots,v_n\}$ be a basis of $V$. If a vector $v$ can be written as


we call $(a_1,a_2,\cdots,a_n)$ the coordinates of $v$ with respect to the basis $B$.

However, I find teaching this definition contradicts our teaching that in a set, the order of elements does not matter. Here if we write $B$ as $\{v_2,\ldots,v_n,v_1\}$, it is still the same set; however, the coordinates of $v$ will usually be different.

So in teaching the coordinates with respect to a basis, should we introduce the notion "ordered basis" (as an ordered $n$-tuple of vectors) and define "the coordinates with respect to an ordered basis" instead? I find that in most Linear Algebra textbooks that I am aware of, this issue is never raised.

  • 4
    $\begingroup$ Yes, a basis is an ordered set. Any book which does not make this clear does not make linear algebra clear. No disagreement here. Actually, I am curious, which book does not make this clear ? $\endgroup$ – James S. Cook Feb 17 '19 at 20:51
  • 2
    $\begingroup$ @JamesS.Cook, I have Serge Lang's Introduction to Linear Algebra by my hand and a basis is defined as merely a set of linearly independent generating vectors. For other books which I have right now, David Lay's Linear Algebra and Ma, Ng, Tan's Linear Algebra have basically the same definition. $\endgroup$ – Zuriel Feb 17 '19 at 20:58
  • 3
    $\begingroup$ This is an issue that has been discussed at length. There are many places in linear algebra where you really want a list rather than a set of vectors. See for example the discussion here: math.stackexchange.com/questions/946477/… $\endgroup$ – Brian Borchers Feb 17 '19 at 21:10
  • 2
    $\begingroup$ Another question which is somewhat related, if $S = \{ v_1, v_2 \}$ is a linearly independent set and $T = S \cup S$ then is $T$ linearly independent ? I would say no. What does that say about my usage of "sets" in linear algebra ? $\endgroup$ – James S. Cook Feb 17 '19 at 21:24
  • 1
    $\begingroup$ @JamesS.Cook, I see your point. I find in Linear Algebra, the usage (by many people) of the word "set" is incorrect, or at least unsatisfactory. $\endgroup$ – Zuriel Feb 17 '19 at 21:25

At the level of standard linear algebra courses, where we consider bases only for finite-dimensional vector spaces, we should work with ordered bases, so that the coefficients of a vector come with an ordering (and thus constitute a vector in $\mathbb R^n$ or $\mathbb C^n$) and so that the matrices that represent linear transformations come with their rows ordered and their columns ordered (and might therefore be written on a page).

At a more abstract level, we could use unordered bases. The coefficients of a vector $v$ would then not be a sequence of scalars but rather a function from the basis into the field of scalars --- the function assigning to each basis element $b$ the coefficient of $b$ in the expansion of $v$. Similarly, transformation matrices would have their rows and columns not ordered but rather indexed by basis vectors. The point of this more abstract approach is that an ordering would be used only to match up which coefficient goes with which basis vector (or, in the case of matrices, which pair of basis vectors), and so we can forget the ordering and work directly with the matching.

I have not tried to teach the more abstract approach in linear algebra classes, and I'm pretty sure I don't want to try. In classes I'll stick with ordered bases.


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.