Lecture Note on Linear Algebra

14. Linear Independence, Bases and

Coordinates
Wei-Shi Zheng,
wszheng@[Link], 2011

November 3, 2011

1 What Do You Learn from This Note

Do you still remember the unit vectors we have introduced in Chapter 1:
     
1 0 0
⃗e1 =  
0 , ⃗e2 =  
1 , ⃗e3 =  0 . (1)
0 0 1
They are a set of vectors which are the basis (specially the standard basis)
of R3 . In this lecture note, we give a more general definition of basis

Basic Concept：Basis(基), standard basis(标准基), coordinate system(坐

标系), coordinate vector(坐标向量)，coordinate mapping(坐标映射)

2 基)
Basis(基
Definition 1 (linear independence). Let V be a vector space. Vectors
⃗v1 , . . . , ⃗vn ∈ V are said to be linearly independent iff (if and only if)
c1⃗v1 + . . . + cn⃗vn = ⃗0
implies
c1 = · · · = cn = 0.
The set {⃗v1 , . . . , ⃗vn } of vectors is called an independent set if ⃗v1 , . . . , ⃗vn are
linearly independent.

1
 注：这里linear independence是一个generalized的定义，是指在一般向
量空间中的定义。如果向量空间是欧式空间，就是我们第一章的线性独立
的定义。在本学期课程，线性空间暂时限制在欧式空间上定义。
Remarks:

1. Vectors ⃗v1 , . . . , ⃗vn are said to be linearly dependent if they are not
linearly independent, or equivalently, there exists c1 , . . . , cn ∈ R which
are not all zero such that

c1⃗v1 + . . . + cn⃗vn = ⃗0.

2. If {⃗v1 , . . . , ⃗vn } is a linearly independent set then any subset of it is

linearly independent.

3. If {⃗v1 , . . . , ⃗vn } is a linearly dependent set then any superset of it, that
is any set contains {⃗v1 , . . . , ⃗vn }, is linearly dependent.

4. As in the case of Rn , ⃗0 itself is linearly dependent. So if one of ⃗v1 , . . . , ⃗vn

is ⃗0, then ⃗v1 , . . . , ⃗vn are linearly dependent.

Definition 2 (basis(基)). Let V be a vector space and B = {⃗b1 , . . . , ⃗bn } an

indexed set of vectors of V . Then B is called a basis of V iff
(1) B is linearly independent;
(2) V = Span B.
Also define the empty set ∅ to be the basis of the trivial space {⃗0}, namely
a vector space contains only the zero vector.

Example: For any n ∈ Z+ , {e1 , . . . , en } is a basis of Rn , which is called the

standard basis (标准基) of Rn .

Example: For any n ∈ Z+ , {1, . . . , xn } forms a basis of Rn [x]. It is obvious

that Rn [x] = Span{1, . . . , xn }. On the other hand, if c0 + c1 x + · · · + cn xn = 0
then c0 = · · · = cn = 0 by equality of polynomials.

2
Theorem 3. Let V be a vector space and S = {⃗v1 , . . . , ⃗vn } ⊆ V . If
V = Span S then some subset of S forms a basis of V .
Proof. 1. If ⃗v1 , . . . , ⃗vn are linearly independent, then by assumption, S is
a basis of V .
2. Otherwise, ⃗v1 , . . . , ⃗vn is linearly dependent. So there exists ⃗vi ∈ S
such that ⃗vi is a linear combination of ⃗v1 , . . . , ⃗vi−1 , ⃗vi+1 , . . . , ⃗vn . De-
fine Sn−1 = {⃗v1 , . . . , ⃗vi−1 , ⃗vi+1 , . . . , ⃗vn }. Then we have V = Span S =
Span Sn−1 . Now Sn−1 ⊆ S is either a basis of V or a linear depen-
dent set. For the case that Sn−1 is linearly dependent, again we can
remove some ⃗vj ∈ Sn−1 from Sn−1 to obtain Sn−2 ⊆ Sn−1 such that
V = Span Sn−2 . Repeat this process and finally, we must obtain a lin-
early independent subset Sk of S such that V = Span Sk . Thus, Sk is
a basis of V .

Theorem 4. If two matrix A and B are row equivalent. Then, the columns
of A has exactly the same linear dependence relationships as the columns of
B.
Proof. As A and B are row equivalent, that is A can be row reduced to a
matrix B. That is the matrix equations A⃗x = ⃗0 and B⃗x = ⃗0 have exactly
the same set of solutions. That is the columns of A have exactly the same
linear dependence relationships as the columns of B.

Theorem 5. The pivot columns of a matrix A form a basis for ColA.

Proof. We prove this theorem by Four steps: (1) We need to use the result
of the last theorem. Let B be the reduced echelon form of A. The set of
pivot columns of B is linearly independent, because no vector in the set is
linear combination of the vectors that precede it.
(2) Since A is row equivalent to B, the pivot columns of A are linearly
independent as well according to Theorem 4.
(3) The nonpivot columns of B must be the linear combination of the pivot
columns of B.
(4) For this same reason, every nonpivot column of A is a linear combination
of the pivot columns of A. Thus the nonpivot columns of A may be discarded
from the spanning set for ColA, by the Spanning Set theorem. This leaves
the pivot columns of A as a basis for ColA.

3
3 Coordinate Systems
We first give the definition of coordinate and then we prove that the
representation is unique.

Definition 6 (coordinate). Suppose B = {⃗b1 , . . . , ⃗bn } is a basis for V and ⃗x is

in V . The coordinates of ⃗x relative to the basis B are  the weights c1 , · · · , cn
c1
⃗ ⃗  .. 
such that ⃗x = c1 b1 + · · · + cn bn . The vector  .  ∈ Rn is called the
cn
coordinate vector of ⃗x with respect to basis B (B–coordinates of ⃗x) and is
denoted by [⃗x]B .

We now prove that the coordinate with respect to basis B is unique.

Theorem 7. Let V be vector space and B ⃗ 1 , . . . , ⃗bn } be a basis of V .

= {b
c1
 .. 
Then for each ⃗x ∈ V , there exists a unique  .  ∈ Rn such that
cn
 
c1
 
⃗x = c1⃗b1 + · · · + cn⃗bn = (⃗b1 · · · ⃗bn )  ...  .
cn

Proof. The existence of (c1 · · · cn )T ∈ Rn follows V = Span B. Now suppose

that (c′1 · · · c′n )T ∈ Rn and ⃗x = c′1⃗b1 + · · · + c′n⃗bn . Then

⃗0 = ⃗x−⃗x = (c1⃗b1 +· · ·+cn⃗bn )−(c′1⃗b1 +· · ·+c′n⃗bn ) = (c1 −c′1 )⃗b1 +· · ·+(cn −c′n )⃗bn .

Since ⃗b1 , . . . , ⃗bn are linearly independent, so c1 − c′1 = · · · = cn − c′n = 0. That

is ci = c′i for all i. The uniqueness follows.

4
4 Coordinate Mapping: A Linear Transfor-
mation View
Given a basis B = {⃗b1 , . . . , ⃗bn } of Rn , for any ⃗x ∈ Rn , its coordinates are
c1 , c2 , · · · , cn . We actually can rewrite the following relations:

⃗x = c1⃗b1 + · · · + cn⃗bn = PB [⃗x]B . (2)

We call PB the change-of-coordinates matrix from B to the standard basis in

Rn .
Note that PB is invertible (because the corresponding linear transforma-
tion is injective and PB is a square matrix). So we also have:

PB−1⃗x = [⃗x]B . (3)

From what we have learned from matrix theory, we have the following:

Theorem 8. Let B = {⃗b1 , . . . , ⃗bn } be a basis of Rn . Then, the linear trans-

formation T : PB−1⃗x → [⃗x]B is injective (one-to-one) from Rn to Rn . We
call PB−1 the Coordinate Mapping.

Now we wish to generalize the above theorem for a more general vector
space V as follows:

Theorem 9. Let B = {⃗b1 , . . . , ⃗bn } be a basis of V . Then, the transformation

T (⃗x) = [⃗x]B is linear and injective (one-to-one) from V to Rn .

Proof. There are two steps to prove this theorem. We first need to prove T
is a linear transformation. Second we need to prove it is one-to-one.
Step 1: Let ⃗u,⃗v be two vectors in V , then we have:

⃗v = c1⃗b1 + · · · + cn⃗bn , ⃗u = d1⃗b1 + · · · + dn⃗bn .

5
So for any scalar e, f , we have

T (e⃗v + f⃗u)
= T ((ec1 + f d1 )⃗b1 + · · · + (ecn + f dn )⃗bn )
 
ec1 + f d1
 .. 
= . 
ecn + f dn
    (4)
c1 d1
   .. 
a = e  ...  + f  . 
cn dn
= e[⃗v ]B + f [⃗u]B
= eT (⃗v ) + f T (⃗u).

Step 2: We need to prove T is one-to-one. Suppose for any ⃗v , ⃗u,

⃗v = c1⃗b1 + · · · + cn⃗bn , ⃗u = d1⃗b1 + · · · + dn⃗bn .

If T (⃗v ) = T (⃗u), then T (⃗v − ⃗u) = ⃗0. That is c1 − d1 = 0, · · · , cn = dn , and

therefore ⃗v − ⃗u

注：下面的概念只要求了解。

Definition 10 (isomorphism(同构)). In general a one-to-one linear transfor-

mation from a vector space V onto a vector space W is called an isomorphism
from V to W

5 Change of Basis
Definition 11 (change–of–coordinates matrix,坐标变换). Let V
be a vector space of dimension n with bases B = {⃗b1 , . . . , ⃗bn } and C =
{⃗c1 , . . . , ⃗cn }. Then for i = 1, . . . , n, we have

⃗ci = (⃗b1 · · · ⃗bn )[⃗ci ]B .

It is convenient to write

(⃗c1 · · · ⃗cn ) = (⃗b1 · · · ⃗bn )([⃗c1 ]B · · · [⃗cn ]B ) = (⃗b1 · · · ⃗bn )[C]B .

The n × n matrix [C]B = ([⃗c1 ]B · · · [⃗cn ]B ) is called the matrix from bases C
to B.

6
Theorem 12. Let V be a vector space of dimension n with bases B =
{⃗b1 , . . . , ⃗bn } and C = {⃗c1 , . . . , ⃗cn }. Then for any ⃗v ∈ V ,

[⃗v ]B = [C]B [⃗v ]C .

Proof. We have

⃗v = (⃗c1 · · · ⃗cn )[⃗v ]C = (⃗b1 · · · ⃗bn )[C]B [⃗v ]C

On the other hand, ⃗v = (⃗b1 · · · ⃗bn )[⃗v ]B . By the uniqueness of coordinates,

we must have [⃗v ]B = [C]B [⃗v ]C .

Remarks: 1. Matrix [C]B is invertible, since its rank is n. It is easy to see

that [C]−1
B = [B]C . So

[⃗v ]C = [C]−1
B [⃗
v ]B = [B]C [⃗v ]B .

2. We know that E = {e1 , . . . , en } is the standard basis of Rn . Now let

B = {⃗b1 , . . . , ⃗bn } be any basis of Rn . Then

[B]E = ([⃗b1 ]E · · · [⃗bn ]E ) = (⃗b1 · · · ⃗bn ).

Examples: Textbook P.274, P.275.

The Feast in the House of Levi, by Veronese