MATH 4A - Linear Algebra With Applications: Lecture 17: Linear Independence, Bases, and Coordinate Systems

Linear independence and bases Whittling down spanning sets Coordinate systems
MATH 4A - Linear Algebra with Applications

Lecture 17: Linear independence, bases, and coordinate systems
10 May 2019
Reading: §4.3-4.6
Recommended problems from §4.3: 1, 3, 5, 7, 9, 11, 13, 15, 21,
22, 23, 33
Recommended problems from §4.4: 1, 3, 5, 9, 11, 13, 15, 16, 17,
19, 21, 27, 31
Announcement: there were some issues with some homework
problems from HW11 and HW12, so I removed them from the
assignments. Also, HW12 includes a few questions from today.
Lecture plan
1 Linear independence and bases
2 Whittling down spanning sets
3 Coordinate systems
Lingering issue
At the moment, our concrete understanding of images and kernels

(among other things!) in terms of matrices only works for linear
transformations Rn → Rm .
We would like to find similarly concrete description for linear

transformations between abstract vector spaces V → W . To this
end, in the next few days we will figure out how to associate
matrices to abstract linear transformations. First, we need to
discuss bases of vector spaces, which generalize the standard basis
vectors of Rn
Yet another not so new definition
A set of vectors {v1 , v2 , . . . , vp } in a vector space V is linearly

independent if the vector equation
c1 v1 + c2 v2 + · · · + cp vp = 0
has only the trivial solution.
In the case V = Rn , this is the exact same definition from before.

We can similarly generalize the definition of linearly dependent and
linear dependence relation from Rn to abstract vector spaces V .
Recall
The standard basis of Rn is the set of n vectors {e1 , e2 , . . . , en }

where      
1 0 0
0 1 0
     
e1 =   , e2 =   · · · en = 0 .
0 0  
. . .
 ..   ..   .. 
0 , 1
This set of vectors has two properties that work especially nicely
together:
1 The vectors span Rn .
2 The vectors are linearly independent.
The first condition says we can generate any vector v in Rn as a
linear combination of the ei , and the second condition implies that
there is exactly one way to write v as such a linear combination.
Key Idea: any set of vectors in a vector space V with these

two properties (called a basis for V ) is “as good as” the
standard basis of Rn
In the next few lectures, we will use such sets of vectors to put
“coordinates” on V , allowing us to identify V with Rn . Then we
can use all of the matrix algebra we know and love to solve
problems in V .
Definition
Let H be a subspace of a vector space V . A basis of H is a set of

vectors B = {b1 , b2 , . . . , bp } in V such that
(i) B is a linearly independent set, and
(ii) H = Span B = Span{b1 , b2 , . . . , bp }.
(Note: condition (ii) forces B to be a subset of H, not just a
subset of V .)
Examples
Note that if V is a vector space, H = V is always a subspace of V .

So it makes sense to talk about a basis of V .
The motivating example is the standard basis of Rn . It is clearly a

basis.
But Rn has lots of other bases. If we choose any n vectors

“randomly” (so that we avoid “probability 0 events” like three
points lying on the same line) then they almost surely span Rn .
E.g. consider the set of vectors
     
 8 1 −2 
B=  −1 , −4 , 1  .
   
−1 0 0
 
Examples
There are many (basically equivalent) ways we could use to

determine if B is a basis. Here’s one way: form the matrix
 
8 −1 −1
A =  1 −4 0 
−2 1 0
and compute its determinant:
det A = 7.
Thus A is invertible, which tells us its column vectors are linearly

independent (because A is one-to-one) and span R3 (because A is
onto).
We conclude B is a basis for R3 .

Examples
Let S = {1, t, t 2 , . . . , t n }. I claim S is a basis for Pn . Why?
If p(t) = c0 + c1 t + c2 t 2 + · · · + cn t n , then clearly
p(t) = c0 · 1 + c1 · t + c2 · t 2 + · · · + cn · t n ,
which shows S spans Pn .
To see S is linearly independent, suppose some linear combination

of the vectors in S satisfies
c0 · 1 + c1 · t + c2 · t 2 + · · · + cn · t n = 0.
From algebra, we know that the only way this is possible is if

c0 = c1 = · · · = cn = 0, which means the only linear dependence
relation on S is the trivial one. Thus S is linearly independent.
From a generating set to a basis
Let v1 , . . . , vp be some vectors in the vector space V . We know
H = Span{v1 , . . . , vp }
is a subspace of V . How can we find a basis for H?

Example
Consider the three vectors

     
19 1 −14
v1 = 13 v2 = −1 v3 = −18 ,
1 4 19
and let H = Span{v1 , v2 , v3 }. Let’s find a basis for H.
Note that v3 = 5 · v2 − v1 (of course,

the way we discover this is
by row reducing v1 v2 v3 0 ). Suppose x is some vector in H.
Then there exist scalars c1 , c2 , c3 such that
x = c1 v1 + c2 v2 + c3 v3 .
Since v3 = 5 · v2 − v1 , we can substitute out the v3 :
x = c 1 v1 + c 2 v2 + c 3 v3
= c1 v1 + c2 v2 + c3 (5 · v2 − v1 )
= (c1 − 5)v1 + (c2 + 5c3 )v2
What does this show? The vector x in H = Span{v1 , v2 , v3 } we

started with is actually in Span{v1 , v2 }!
Of course, by definition, Span{v1 , v2 } is a subspace of H. What

we just showed is the converse: H is a subspace of Span{v1 , v2 }.
We conclude that H = Span{v1 , v2 }.
Since v1 and v2 are not scalar multiples of one another, they are
linearly independent.
We conclude that {v1 , v2 } is a basis for H.

More generally
When we consider H = Span{v1 , . . . , vp }, any nontrivial linear
dependence relation among the v1 , . . . , vp allows us to remove a
vector. More precisely:
Theorem
Let S = {v1 , . . . , vp } be a set in V , and let H = Span{v1 , . . . , vp }.
(a) If one of the vectors in S—say, vk —is a linear combination of

the remaining vectors in S, then the set formed from S by
removing vk still spans H.
(b) If H 6= {0}, some subset of S is a basis of H.
Intuitively: if S is a spanning set of H, then we can toss out some

of the vectors in S to get a basis of H.
(The proof of part (a) is a simple generalization of the example

above. To prove part (b), apply part (a) several times.)
Application: finding a basis for column spaces
Let A be a m × n matrix. We can apply the previous theorem to

find a basis for the column space Col A.
By definition, Col A is the subspace of Rm spanned by the columns

of A. So the previous theorem tells us some subset of the columns
of A is a basis of Col A.
Which subset?
Application: finding a basis for column spaces
Theorem
The pivot columns of a matrix A form a basis for Col A.
NOTE: don’t forget that the pivot columns of A are the columns
of A that contain pivots after reducing to echelon form E . They
are NOT columns of E .
(Proof: Let E be an echelon form of A. It is easy to see the pivot

columns of E are linearly independent. The set of solutions to
E x = 0 are the same as the set of solutions to Ax = 0, and
solutions to Ax = 0 are the same thing as linear dependence
relations between the columns of A. Thus, the pivot columns of A
are linearly independent. By similar reasoning, the nonpivot
columns of A are linear combinations of the pivot columns, so we
don’t need them when we compute Col A.)
A little bit of a tangent: finding a basis for a null space
In the previous lecture, we discussed how to find a spanning set of

the null space of a matrix A: row reduce the augmented matrix
(A 0) to determine the parametric vector form of the solution set.
In fact, this procedure determines a basis of Nul A. I’ll let you
think about that.
Motivation
Bases are the most important way to make abstract vector spaces
V concrete.
The basic idea (pun intended) is to “identity” each of the vectors

bi in a basis {b1 , . . . , bn } of V with the standard basic vector ei in
Rn . Let’s make this more precise.
A preliminary uniqueness result
Theorem
Let B = {b1 , . . . , bn } be a basis for the vector space V . Then for
each vector x in V , there exists a unique set of scalars c1 , . . . , cn
such that
v = c1 b1 + · · · + cn bn .
Definition
Let B = {b1 , . . . , bn } be a basis for the vector space V . The

coordinates of x relative to B are the weights c1 , . . . , cn such that
x = c1 b1 + · · · + cn bn .
The coordinate mapping determined by B is the linear

transformation (we’ll prove it’s linear shortly)
V → Rn
 
c1
 c2 
x 7→ [x]B =  . 
 
 .. 
cn
We call [x]B the coordinate vector of x determined by B.

Example
Consider the basis B = {1, t, t 2 , t 3 } of P3 . If
p(t) = a0 + a1 t + a2 t 2 + a3 t 3
is a vector in P3 , its coordinate vector with respect to B is

 
a0
a1 
[p(t)]B = 
a2  .

a3
Example
Consider the nonstandard basis of R2

−3 −1
B= ,
13 1
and the vector

1
x= .
2
To determine [x]B , we must figure out how to write x as a linear
combination of the vectors in B.
In other words, we must solve the linear system

−3 −1 1
13 1 2
The reduced echelon form is

3

1 0 10
0 1 − 19
10
Thus
3 −3 19 −1
x= − ,
10 13 10 1
so
3/10
[x]B = .
−19/10
iClicker
Consider the basis

−1 1
B= ,
1 1
of R2 , and the vector
0
x= .
1
What is the first entry of [x]B ?
(a) 1
(b) −1
(c) 1/2
√
(d) 2/2
(e) −1/2
The coordinate mapping
Theorem
Let B be a basis for the vector space V . Then the coordinate
mapping is a one-to-one and onto linear transformation.

MATH 4A - Linear Algebra With Applications: Lecture 17: Linear Independence, Bases, and Coordinate Systems

Uploaded by

Copyright:

Available Formats

MATH 4A - Linear Algebra With Applications: Lecture 17: Linear Independence, Bases, and Coordinate Systems

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MATH 4A - Linear Algebra With Applications: Lecture 17: Linear Independence, Bases, and Coordinate Systems

Uploaded by

Copyright:

Available Formats

Linear independence and bases Whittling down spanning sets Coordinate systems

MATH 4A - Linear Algebra with Applications

1 Linear independence and bases

2 Whittling down spanning sets

At the moment, our concrete understanding of images and kernels

We would like to find similarly concrete description for linear

Yet another not so new definition

A set of vectors {v1 , v2 , . . . , vp } in a vector space V is linearly

has only the trivial solution.

In the case V = Rn , this is the exact same definition from before.

The standard basis of Rn is the set of n vectors {e1 , e2 , . . . , en }

Key Idea: any set of vectors in a vector space V with these

Let H be a subspace of a vector space V . A basis of H is a set of

Note that if V is a vector space, H = V is always a subspace of V .

The motivating example is the standard basis of Rn . It is clearly a

But Rn has lots of other bases. If we choose any n vectors

There are many (basically equivalent) ways we could use to

and compute its determinant:

Thus A is invertible, which tells us its column vectors are linearly

We conclude B is a basis for R3 .

Let S = {1, t, t 2 , . . . , t n }. I claim S is a basis for Pn . Why?

If p(t) = c0 + c1 t + c2 t 2 + · · · + cn t n , then clearly

which shows S spans Pn .

To see S is linearly independent, suppose some linear combination

From algebra, we know that the only way this is possible is if

From a generating set to a basis

Let v1 , . . . , vp be some vectors in the vector space V . We know

is a subspace of V . How can we find a basis for H?

Consider the three vectors

and let H = Span{v1 , v2 , v3 }. Let’s find a basis for H.

Note that v3 = 5 · v2 − v1 (of course,

Since v3 = 5 · v2 − v1 , we can substitute out the v3 :

What does this show? The vector x in H = Span{v1 , v2 , v3 } we

Of course, by definition, Span{v1 , v2 } is a subspace of H. What

We conclude that {v1 , v2 } is a basis for H.

(a) If one of the vectors in S—say, vk —is a linear combination of

Intuitively: if S is a spanning set of H, then we can toss out some

(The proof of part (a) is a simple generalization of the example

Application: finding a basis for column spaces

Let A be a m × n matrix. We can apply the previous theorem to

By definition, Col A is the subspace of Rm spanned by the columns

Application: finding a basis for column spaces

(Proof: Let E be an echelon form of A. It is easy to see the pivot

A little bit of a tangent: finding a basis for a null space

In the previous lecture, we discussed how to find a spanning set of

The basic idea (pun intended) is to “identity” each of the vectors

A preliminary uniqueness result

Let B = {b1 , . . . , bn } be a basis for the vector space V . The

The coordinate mapping determined by B is the linear

We call [x]B the coordinate vector of x determined by B.

Consider the basis B = {1, t, t 2 , t 3 } of P3 . If

is a vector in P3 , its coordinate vector with respect to B is

Consider the nonstandard basis of R2

and the vector  

In other words, we must solve the linear system

The reduced echelon form is

Consider the basis    

The coordinate mapping

and the vector

Consider the basis