Linear Algebra

Lecture 5: Linear transformations and matrices.

Jan-Otto Hooghoudt

Aalborg University (AAU)

Esbjerg, 2019
Linear Dependency:

A set of k
vectors {ū1 , ū2 , . . . , ūk } in R n is called linearly
dependent if
there exist scalars c1 , c2 , . . . , ck , not all equal to 0, such that
c1 ū1 + c2 ū2 + . . . + ck ūk = 0̄

or equivalent: if and only if there exist a nonzero solution

of Ax̄ = 0̄ where A = [ū1 ū2 . . . ūk ]
Rep: Linear Independency

A set of vectors {ū1 , ū2 , . . . , ūk } in R n is called linearly

k
independent if they are not linear dependent
implying that: The set {ū1 , ū2 , . . . , ūk } in R n is Linearly
Independent if and only if the only solution of Ax̄ = 0̄
where A = [ū1 ū2 . . . ūk ] is the zero solution i.e. x̄ = 0̄.
Rep: Equivalence relations in relation to: the columns of A
are linearly independent.
The following statements about an m × n matrix A are equivalent:
Theorem
a) The columns of A are linearly independent.
b) The equation Ax̄ = b̄ has at most one solution for each b̄ in
Rm
c) The nullity of A is zero.
d) The rank of A is n, the number of columns of A.
e) The columns of the reduced row echelon form of A are distinct
standard vectors in R m .
f) The only solution of Ax̄ = 0̄ is 0̄.
g) There is a pivot position in each column of A.
Proof:
Linear transformations and matrices
Denition: (Vector-)function; Image, (Co-)domain and range
Denition: Let S and S be subsets of R n and R m , respectively.
1 2

A function f from S1 to S2 , written as f : S1 → S2 , is a rule

that assigns to each vector v̄ in S 1 a unique vector f (v̄ ) in S2 .

the vector f (v̄ ) is called the image of v̄ (under f )

the set S 1 is called the domain of a function f ,
while the set S 2 is called the codomain of f .
the range of f is dened to be the set of images f (v̄ ) for all v̄
in S1 .
Linear transformations and matrices

Denition: Function dened through a Matrix multiplication

Let A be an m × n matrix.
The function TA : R n → R m dened by TA (x̄ ) = Ax̄ for all x̄ in
R n is called: the matrix transformation induced by A.
Observe that there is a reversal in the order of the size of the
matrix m × n and the dimension of the domain R n and the
codomain R m of TA (!)
Linear (matrix) transformations: The Rotation matrix

Example: One important example of a matrix transformation is

e.g the in Section 1.2 introduced rotation matrix Aθ

That is TAθ : R 2 → R 2 represents the function that rotates a

vector counterclockwise by θ .
Linear transformations and matrices: projection matrix

Another example of a matrix transformation is e.g. dened through

a projection matrix.
1 0 0
 

Example: Let A be the matrix 0 1 0 ,

0 0 0
then TA : R → R is dened by
3 3

 x1  1 0 0 x1 x1
      

TA x2  = 0 1 0 x2  = x2 

x3 0 0 0 x3 0

Obviously, TA (ū ) is the orthogonal projection of ū on the xy -plane,

and the range of TAθ is the xy -plane in R 3
Linear transformations and matrices: shear matrix
1 k
 
Let k be a scalar and A = , then the function
0 1
 x   x + kx 

TA : R → R is dened by TA x1 = 1 x 2 and is called
2 2
2 2
a shear transformation.

Notice the eect

on the vector ū in the upper gure;
the head of the vector is moved to
the right, but stays at the same
height.
on the the letter I centered on the
y-axis in the lower gure, where in
this case k = 2.
Linear transformations and matrices: other examples

Other important linear geometric transformations are:

reections
contractions and
dilations (the opposite of a contraction)
Denition: Linear Transformation(!)

Denition: Linear Transformation

A function T : R n → R m is called a linear transformation (or
simply linear) if, for all vectors ū and v̄ in R n and all scalars c ,
both of the following conditions hold:
(i) T (ū + v̄ ) = T (ū ) + T (v̄ )
(ii) T (c ū ) = cT (ū )
Every matrix transformation is linear(!)

For any m × n matrix A and any vectors ū and v̄ in R n , the

following statements are true:
(a) TA (ū + v̄ ) = TA (ū ) + TA (v̄ )
(b) TA (c ū ) = cTA (ū ) for every scalar c
Both follow from Theorem 1.3 on page 24, which states properties of the
Matrix-Vector product

That is: Every transformation dened by a matrix is linear(!)

Two special linear transformations

We introduce:
the identity transformation I : R n → R n , which is dened
by I (x̄ ) = x̄ for all x̄ in R n
the zero transformation T0 : R n → R m , which is dened by
T0 (x̄ ) = 0̄ for all x̄ in R n , and has range 0̄ ∈ R m
Both are indeed linear transformations, as can be veried by
checking the denition.
Basic properties of linear transformations

Theorem: Basic properties of linear transformations

For any linear transformation T : R n → R m , the following
statements are true:
(a) T (0̄) = 0̄
(b) T (−ū ) = −T (ū ) for all vectors ū in R n
(c) T (ū − v̄ ) = T (ū ) − T (v̄ ) for all vectors ū and v̄ in R n
(d) T (aū + bv̄ )= aT (ū ) + bT (v̄ ) for all vectors ū and v̄ in R n
and all scalars a and b.
Example 5 book
 1  2
   1   4 
Suppose that T1
=
3
and T −1 = −1
 3 
 
(a) Find T
3
 1   0 
(b) Find T and T , and use the results to determine
0 1
T x1 .
   
x2
Solution:
3 1
   
(a) Because =3 , it follows immediately that:
3 1
T 33 = T 3 11 = 3T 11 = 3 23 = 69
          
Example 5 book: continued
 1 
(b) The question is to nd T . Therefore we try to nd rst
0
1 1 1
     
the linear combination of and that gives , that is
1 −1 0
a +b
a 1 + b −1 = 0 =⇒ a − b = 10 =⇒ a = b =
1 1 1 1
         

2
Thereby:
 1   1 1 1  1  
T 0 =T +
2 1 2 −1
1 1 1  1 
   
= T + T

2 1 2 −1
1 2 1 4 3
     
= + =
2 3 2 −1 1
Example 5 book: continued

(b-continued)
 Similarly, we try
 to nd the linear combination of
1 1 0
 
and that gives , that is
1 −1 1

a 11 + b −11 = 01 =⇒ aa + b = 0 =⇒ b = − 1 , a = 1
         

−b 1 2 2
Thereby:
 0   1 1−1 1 
 
T 1
=T +
2 1 2 −1
1 1 1  1 
   
= T − T

2 1 2 −1
1 2 1 4 −1
     
= − =
2 3 2 −1 2
Example 5 book: continued

 x  
Now we use the 2 previous result obtained to determine T x2
1

by:
 x   1
 
0
 
T = T x1 + x2

1
x2 0 1
 1   0 
= x1 T + x2 T
0 1
3 −1 3x1 − x2
     
= x1 + x2 =
1 2 x1 + 2x2
Every linear transformation is a matrix transformation(!!)

The previous example was an example of validity of the following theorem that
is of central importance within the eld of Linear Algebra:
Theorem
Let T : R n → R m be linear. Then there is a unique m × n matrix
A = [T e1 T e2 . . . T en ]

whose columns are the images under T of the standard vectors for R n , such
that T (v̄ ) = Av̄ for all v̄ in R n .

That is every linear transformation with domain R n and codomain R m is a

matrix transformation(!!).
Which means that if a transformation T is linear, we can produce a
corresponding matrix A such that the transform T is given by transform
dened through the matrix A and we write T = TA .
Proof of previous Theorem:
Note: In the 1st step of the proof we use that it (trivially) holds that:
v1
 
 v2 
v̄ =  .  = v1 e1 + v2 e2 + . . . + vn en
 
 .. 
vn
for anyv̄ in Rn . Then:
T (v̄ ) = T (v1 e1 + v2 e2 + . . . + vn en ) (1)
= v1 T (e1 ) + v2 T (e2 ) + . . . + vn T (en ) (2)
= Av̄ (3)
= TA (v̄ ) (4)
-From (1) to (2) as T is linear;
-From (2) to (3)by denition of matrix-vector product and dening
A = [T (e1 ) T (e2 ) . . . T (en )]
-From (3) to (4): By the denition that: TA (v̄ ) means the linear transformation
dened by the matrix A, where in this case A = [T (e1 ) T (e2 ) = a2 . . . T (en )]
So for a genaral transformation T (v ) we have shown that it can be dened
by the matrix A related to T by: A = [T (e1 ) T (e2 ) = a2 . . . T (en )]. !!
The standard matrix of T

Let T : R n → R m be a linear transformation. We call the m×n

matrix

A = [T (e1 )T (e2 ) . . . T (en )]

the standard matrix of T , with the property that T (v̄ ) = Av̄ for
every v̄ in R n .
Range of linear transformation

We have seen that each linear transformation T : R n → R m can be

dened by dening the corresponding. m × n matrix

A = [T e1 T e2 . . . T en ]

such that T (v̄ ) = Av̄ for all v̄ in R n.

Thereby we can observe that:

The range of a linear transformation equals the span of the

columns of its standard matrix.
An onto-function

Denition:
A function f : R n → R m is said to be onto if its range is all of
R m that is, if every vector in R m is an image.

From the previous slide and the denition above, it follows that a
linear transformation is onto if and only if the columns of its
standard matrix form a generating set for its codomain(!)
Equavalent conditions for a linear function to be "onto"

Theorem 2.10
Let T : R n → R m be a linear transformation with standard matrix
A. The following conditions are equivalent:
(a) T is onto; that is, the range of T is R m
(b) The columns of A form a generating set for R m
(c) rank A=m
A one-to-one function

Denition: One-to-one function

A function f : R n → R m is said to be one-to-one if every pair of
distinct vectors in R n has distinct images.
That is, if ū and v̄ are distinct vectors in R n , then f (ū ) and
f (v̄ ) are distinct vectors in R m .
Solution to Ax̄ = 0 for a linear One-to-one function

Suppose that T : R n → R m is a one-to-one linear

transformation. Then, from the one-to-one-property and as
T (0̄) = 0̄ we nd that:
the 0̄ vector is the only vector in R n whose image under T is
the zero vector of R m

Conversely, if 0̄ is the only vector whose image under T is the

zero vector, then T must be one-to-one.
Proof: for suppose that ū and v̄ are vectors with T (ū ) = T (v̄ ) =⇒
T (ū ) − T (v̄ ) = 0̄ =⇒ T (ū − v̄ ) = 0̄, but as we assume that 0̄ is the
only solution to Ax̄ = 0̄ then (ū − v̄ ) = 0̄ =⇒ ū = v̄ , and hence T is
one-to-one.
Connection null space with a One-to-one function

Null space
Denition Let T : R n → R m be linear. The null space of T is the
set of all v̄ in R n such that T (v̄ ) = 0̄.

From the previous denition and the previous slide we can state:
Connection Null-space and one-to-one-function
A linear transformation is one-to-one if and only if its null
space contains only 0̄. With A the standard matrix of the linear
transformation T , then the null space of T is the set of solutions of
Ax̄ = 0̄.
A one-to-one function: Equivalent relations

Theorem 2.11:
Let T : R n → R m be a linear transformation with standard matrix
A. Then the following statements are equivalent:
(a) T is one-to-one.
(b) The null space of T consists only of the zero vector.
(c) The columns of A are linearly independent.
(d) rank(A) = n.
Self Study

Section 2.7:
Example 1, 6,7 and 8
Section 2.8:
Example 14 (pages 181185)

-Further, please read the note "To the student" (if you did not do it yet)
and the general preface (both from the book compiled by O. Geil), that I
added on Moodle.
-Install Matlab in order to be ready for the Self-Study session in 2 days