0% found this document useful (0 votes)
6 views27 pages

Linear Algebra: Vector Spaces & Bases

Lecture 3 of MA1513 covers key concepts in Linear Algebra including vector spaces, subspaces, linear independence, basis, and dimension. It uses color mixing as an analogy to explain linear combinations and discusses the properties of solution spaces for homogeneous and non-homogeneous systems. The lecture also emphasizes the importance of identifying linearly independent sets and bases for vector spaces.
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
6 views27 pages

Linear Algebra: Vector Spaces & Bases

Lecture 3 of MA1513 covers key concepts in Linear Algebra including vector spaces, subspaces, linear independence, basis, and dimension. It uses color mixing as an analogy to explain linear combinations and discusses the properties of solution spaces for homogeneous and non-homogeneous systems. The lecture also emphasizes the importance of identifying linearly independent sets and bases for vector spaces.
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

MA1513

Linear Algebra with


Differential Equations
LECTURE 3
Log in to [Link]/vtpoll using your NUS ID

MA1513 LECTURE 3 1
Topics for week 3
Chapter 2: Vector Spaces
2.3 Subspaces
2.4 Linear independence
2.5 Basis and Dimension
2.6 Coordinate vectors
2.7 Projection and Linear approximation

MA1513 LECTURE 3 2
Color Mixing (An analogy)
Three primary colors: Red, Green, Blue (RGB)
Different color shade combination gives “all” colors

e.g. 20% Red + 45% Green + 30% Blue


this specific linear combination of RGB will produce a particular color

The three primary colors span the entire color space:


span{Red, Green, Blue, Yellow} = Color space
- span{Red, Green, Blue} = Color space
{Red, Green, Blue}
None of the three primary colors are redundant: forms a basis for
Red, Green, Blue, Yellow is not linearly independent the color space
- {Red, Green, Blue} is linearly independent
MA1513 LECTURE 3 3
Overview
All the linear combinations
R3 c1u1 + c2u2 + c3u3
span
{u1, u2, u3} span{u1, u2, u3} We say: span{u1, u2, u3}
assume: coplanar is a subspace of R3

{u1, u2, u3} is a building


block of span{u1, u2, u3}

{u1, u2, u3} is not a basis


for span{u1, u2, u3}
MA1513 LECTURE 3 4
Overview
All the linear combinations
R3 c1u1 + c2u2
span
{u1, u2} span{u1, u2} We say: span{u1, u2}
is a subspace of R3

{u1, u2} is a basis


for span{u1, u2}

We say the dimension


of span{u1, u2} is 2

MA1513 LECTURE 3 5
Overview
span{u1, u2, u3} span{u1, u2}

All the linear combinations All the linear combinations


c1u1 + c2u2 + c3u3 c1u1 + c2u2

There is a redundant vector There is no redundant


among {u1, u2, u3} vector among {u1, u2}
{u1, u2, u3} is a linearly dependent set {u1, u2} is a linearly independent set

MA1513 LECTURE 3 6
Overview
All the linear combinations
R3 span c1u1 + c2u2 + c4u4
{u1, u2, u4} span{u1, u2, u4} = R3
span{u1, u2, u4}
assume: non-coplanar
span{u1, u2, u4} is
again a subspace of R3

{u1, u2, u4} is a linearly


independent set
{u1, u2, u4} is a basis for We say the dimension
span{u1, u2, u4} of span{u1, u2, u4} is 3
{u1, u2, u4} is a basis for R3

MA1513 LECTURE 3 7
Extending to Rn

Rn • Start with a set S of n-vectors {u1, u2, …, uk} in Rn


• span(S) is a subspace of Rn
span(S)
• {u1, u2, …, uk} is a building block of span(S)
c1u1 + c2u2 + ··· + ckuk
• If {u1, u2, …, uk} is linearly independent, it is a basis
for span(S) and the dimension of span(S) is k.

S
u1
u2

uk

MA1513 LECTURE 3 8
In general, number of
dim(solution space) = free parameters
in the general solution

Solution Space of Ax = 0
 2v + 2w − x + z = 0
homogeneous  − v − w + 2 x − 3y
 + z = 0
system 
 x + y + z = 0
Gaussian
 v + w − 2 x − z = 0
Elimination
v  − s − t 
general w   s 
solution x =  
y   − t 
separate    0 
parameters z  t 
 − 1  − 1
general 1 0
linear combination
 
= s 0 + t  − 1
0 0 basis for solution space = {u1, u2}
of vectors    
extract the 0 1
dim(solution space) = 2
vectors u1 u2
the vectors obtained linear span span{u1, u2} = solution space of Ax = 0
is always linearly of the vectors subspace of R5
independent
MA1513 LECTURE 3 9
All subspaces of Rn satisfy closure properties

Solution set of linear system Rn


Solutions of
Ax = b
Ax = b linear system with n variables

system is homogeneous system is non-homogeneous


Ax = 0 Ax = b
The set of solutions is The set of solutions is
a “span”, and hence a not a “span”, and hence
subspace of Rn not a subspace of Rn
We call it a solution space We don’t call it a solution space
Satisfies closure properties Does not satisfy closure properties

If u, v are solutions of Ax = 0, If u, v are solutions of Ax = b,


• u + v also a solution • u + v not a solution
• cu also a solution • cu not a solution

MA1513 LECTURE 3 10
Linearly independent set: No redundant vectors in the set
Linearly dependent set: redundant vectors in the set

Example 1 (Linear independence)


Are these sets of vectors in R3 linearly independent?

{ (1, 1, 1), (0, 1, 1), (0, 0, 1) }

{ (1, 1, 1), (0, 1, 0), (1, 0, 1) }

{ (1, 1, 1), (-1, -1, -1) }

{ (1, 1, 1), (1, 1, -1) }

{ (1, 1, 1), (0, 1, 1), (0, 0, 1), (1, 0, 1) }

MA1513 LECTURE 3 11
Example 1 (cont.)
{ (1, 1, 1), (0, 1, 1), (0, 0, 1) }

Standard method: a(1, 1, 1) + b(0, 1, 1) + c(0, 0, 1) = (0, 0, 0)

1 0 0 1 0 00
G.E  REF 
Alternative method: 1 1 0 =1 1 1 00
only trivial solution  lin indep
1 1 1 1 1 10

{ (1, 1, 1), (0, 1, 0), (1, 0, 1) }

Standard method: a(1, 1, 1) + b(0, 1, 0) + c(1, 0, 1) = (0, 0, 0)

1 0 1 1 0 10
G.E  REF 
Alternative method: 1 1 0 =0 1 1 00
non-trivial solution  lin dep
1 0 1 1 0 10
MA1513 LECTURE 3 12
Example 1 (cont.)
{ (1, 1, 1), (-1, -1, -1) } the set has only two vectors
Standard method
Short-cut method scalar multiples of each other  linearly dependent
{ (1, 1, 1), (1, 1, -1) } the set has only two vectors
Standard method
Short-cut method not scalar multiples of each other  linearly independent
{ (1, 1, 1), (0, 1, 1), (0, 0, 1), (1, 0, 1) } the set has more than 3 vectors in R3
Standard method
Short-cut method more than 3 vectors in R3  linearly dependent

MA1513 LECTURE 3 13
Testing Linear dependency { v1, v2, …, vn }
Standard Method:
• Form the equation c1v1 + c2v2 + … + cnvn = 0.
Check whether the homogeneous system has non-trivial solutions for
c1 , c2 , … cn .
Alternative Methods: (only work under certain circumstances) If Yes,
then linearly dependent
• Use the column vectors v1, v2, …, vn to form a matrix A.
If No,
If A is a square matrix, check whether det(A) = 0. then linearly independent
• If there are only two vectors v1, v2 in the set,
check that v1, v2 are scalar multiple of each other.
• If v1, v2, …, vn ∈ Rm and n > m,
If n < m, no conclusion
then the set is linearly dependent.

MA1513 LECTURE 3 14
Example 1 revisited (Basis)
Which of the sets of vectors is/are bases for R3?

{ (1, 1, 1), (0, 1, 1), (0, 0, 1) }

{ (1, 1, 1), (0, 1, 0), (1, 0, 1) }

{ (1, 1, 1), (-1, -1, -1) }

{ (1, 1, 1), (1, 1, -1) }

{ (1, 1, 1), (0, 1, 1), (0, 0, 1), (1, 0, 1) }

Any set of n linearly independent vectors is a basis for Rn

MA1513 LECTURE 3 15
Announcement
• Online quiz
o Quiz 1: view the correct answers (from Canvas > Quizzes)
o Quiz 2: closing on Sunday night.
Note: If you start a new attempt for a quiz
o Quiz 3: opened
without completing, it will be counted as
your last attempt.
• Grades in Canvas
o Lecture quiz scores (Lecture Quiz 2)
o Group discussion (Tutorial) scores (GD 1)
o Online quiz scores (Quiz 1)
o Practice scores (Practice 1, when it is ready)

MA1513 LECTURE 3 18
G.E. preserves
linear dependency of the columns

Example 2 (Basis for linear span)


need to remove the redundant
V = span{ u1, u2, …, uk } vectors among u1, u2, …, uk

span{(1,2,2,1), (3,6,6,3), (4,9,9,5), (-2,-1,-1,1), (5,8,9,4)}


Form a matrix using u1, u2, …, uk as its columns

1 3 4 −2 5 1 3 4 −2 5
2 6 9 −1 8 G.E 0 0 1 3 −2 REF
2 6 9 −1 9 0 0 0 0 1
1 3 5 1 4 0 0 0 0 0
take columns corresponding pivot columns are linearly independent
to pivot columns of REF non-pivot columns are redundant
Basis for the linear span:
{(1,2,2,1), (3,6,6,3) , (4,9,9,5), (-2,-1,-1,1) , (5,8,9,4)}
MA1513 LECTURE 3 19
Subspace in MA1513
• Rn

• span{v1, v2, …, vk} where v1, v2, …, vk are vectors in Rn subspace of Rn

• Solution space of a homogeneous system Ax = 0 with n variables subspace of Rn

• Row space of an m x n matrix A subspace of Rn

• Column space of an m x n matrix A subspace of Rm

• Nullspace of an m x n matrix A subspace of Rn

• Eigenspace of an n x n matrix A (Chapter 3) subspace of Rn

MA1513 LECTURE 3 21
Row Space and Column Space of A
 1 2 2 1  r1
 
 3 6 6 3  r2
 4 9 9 5  r3
A=  
− 2 − 1 − 1 1  r4
 
 5 8 9 4  r5
 4
 2 7 3  r6
c1 c2 c3 c4
row space of A = span{r1 , r2 , r3 , r4 , r5 , r6 } subspace of R4
column space of A = span{c1 , c2 , c3 , c4 } subspace of R6

MA1513 LECTURE 3 22
All the vectors v in Rn

Nullspace of A that are “nullified” by A,


i.e. Av = 0.

 1 2 2 1
  This is same as the solutions of Ax = 0
 3 6 6 3
 4 9 9 5 Nullspace of A = solution space of Ax = 0
A=  
− 2 − 1 − 1 1 So it is a subspace of Rn
 
 5 8 9 4
 4
 2 7 3 

Form the homogeneous system Ax = 0


and solve (using G.E.) as before.

MA1513 LECTURE 3 23
G.E. preserves
linear dependency of the columns

Basis for Column Space


pivot columns
 1 2 2 1 1 2 2 1
   
 3 6 6 3 0 1 1 1
 4 9 9 5 Gaussian 0 0 1 1
A=   →
  
− 2 − 1 − 1 1 Elimination 0 0 0 0
   
 5 8 9 4 0 0 0 0
0 0 0 0 
 4
 2 7 3  
c1 c2 c3 c4 REF
column space of A = span{c1 , c2 , c3 , c4 }
Basis for column space of A = columns in A corresp to pivot columns of REF
={c1 , c2 , c3 }
Dimension of column space = # pivot columns of REF = rank(A) = 3
MA1513 LECTURE 3 24
G.E. does not preserve linear dependency of the rows
G.E. preserves the row spaces

Basis for Row Space dim(row space) = dim(column space)

 1 2 2 1  r1 1 2 2 1  v1
   
 3 6 6 3  r2 0 1 1 1  v2
 4 9 9 5  r3 Gaussian 0 0 1 1  v3
A=   →
  
− 2 − 1 − 1 1  r4 Elimination 0 0 0 0
   
 5 8 9 4  r5 0 0 0 0
0 0 0 0 
 4
 2 7 3  r6 
REF
row space of A = span{r1 , r2 , r3 , r4 , r5 , r6 } = row space of REF = span{v1 , v2 , v3}
Basis for row space of A = non-zero rows in REF ={v1 , v2 , v3 }
Dimension of row space = # nonzero rows of REF = rank(A) = 3

MA1513 LECTURE 3 25
Dimension Theorem of Matrix
rank(A) + nullity(A) = # columns of A

Basis for Nullspace


Nullspace of Solution space of
matrix A system Ax = 0

Basis for nullspace Basis for solution space

Dimension of nullspace Dimension of solution space

Nullity of A # parameters in general solution


# non-pivot columns of REF

MA1513 LECTURE 3 26
Least Squares Solutions
When a linear system Ax = b is inconsistent

A least squares solution x0 of Ax = b :

• x0 is the best approximation to a solution of Ax = b


Ax0 ≈ b ||Ax0 - b||
closest smallest

How to find the least squares solution x0 of Ax = b?


 Multiply AT on both sides of Ax = b
 Solve the new system ATAx = ATb

MA1513 LECTURE 3 28
Example 3 (Least Squares)
2x + y = 20 2 1 20 2 6 20 30 40
6x + y = 18 6 1 x 18 1 1 1 1 1
2940 98 x0 608
20x + y = 10 20 1 y = 10 98 5 y0 =
56
30x + y = 6 30 1 6 AT
40x + y = 2 40 1 2 ATA ATb
A b
Find least squares solution x0 for Ax = b x0 is the actual solution of ATAx = ATb

x0 2940 98
−1
608 −206/637
What if ATA is singular? x0 = y = =
0 98 5 56 268/13
Solve the new system using GE
(ATA)-1
2940 98 608
98 5 56

MA1513 LECTURE 3 29
If x0 is the least squares solution of Ax = b,
Ax0 = projection of b onto column space of A Rn
b
U p
Projection on Subspaces u1 u 2 … u k

Given a subspace U of Rn and a vector b not in U


Find a vector p in U “nearest” to b How to find this projection of b?
This vector p is called
the projection of b onto a subspace U

 Find any basis {u1, … , uk} for U  span{u1, … , uk} = U


 Form matrix A using u1, … , uk as column vectors  span{u1, … , uk}
= column space of A
 Consider the linear system Ax = b
 Solve the system ATAx = ATb, say x0 least squares solution of Ax = b
 projection p = Ax0

MA1513 LECTURE 3 30
If x0 is the least squares solution of Ax = b, (1,-1,1)
Ax0 = projection of b onto column space of A U x+y+z=0

Example 4 (Projection)
Find the projection of (1,-1,1) onto the plane U: x + y + z = 0 in R3
U = span{ 2 non-parallel vectors on the plane } = span{ (1,-1,0), (1,0,-1) }
1 1 1
Form matrix A = −1 0 and column vector b = −1
0 −1 1

Find the least squares solution of Ax = b


2 1 x = 2 x 4/3
Solve A Ax = A b
T T =
1 2 y 0 y −2/3
1 1 2/3
Projection of (1, -1, 1) on U: 4/3
−1 0 = −4/3
−2/3
0 −1 2/3

MA1513 LECTURE 3 31

You might also like