001 - Lecture - 3 - Annotated - Matrix Multiplication

Matrix multiplication
 Suppose A and B can be multiplied to form a matrix C

n
 Then AB  C , where cik  a
m 1
b
im mk
 cik is the inner product of

 the i-th row from A and
 the k-th column from B
Matrix multiplication: Example
 Ex:  a11 a12   b11 b12   c11 c12 

a  b   
 21 a 22   21 22   21 22 
b c c
Matrix multiplication 2
 cik is defined only if the 2 vectors have the same
dimension
==>
 Conformability: A and B can be multiplied to form AB
only if
column dimension of A = row dimension of B
 Ex: b11 b12 

a11 a12 a13 
b21 b22 
a a  
 21 22 a23 
b31 b32 
Matrix multiplication 3
 If
A is of dimension m x n
B is of dimension n x p
==> C = AB is an m x p matrix
Plan for Today
 Matrix multiplication – examples
 Matrices vs vectors
 Special matrices:
 Identity matrix
 Null matrix
 Idempotent matrix
 Rules of matrix algebra
 Transpose of a matrix?
Matrix multiplication - Example
C = AB =
Matrix multiplication – Example
BA =
Matrix multiplication - Example
 Note: AB and BA are different things!

 in AB, we pre-multiply B by A
 in BA, we post-multiply B by A
Matrices and Systems of Equations
 Back to
Matrix multiplication - Matrices & Vectors
 Recall: Vectors can be viewed as special types of

matrices
Matrix multiplication - Matrices & Vectors
 However: If x is a column matrix and a y’ is a row

matrix, then matrix product xy’ ≠ inner product xꞏy’ !
1 
 Ex 1: x    y '  3 5
 4
x∙y’ =
xy’ =
Matrix multiplication - Matrices & Vectors 2
1 
 Ex 2: u  v'  3 5 2
 4
u∙v’ =
uv’ =
Matrix multiplication - Matrices & Vectors 3
 Ex 3: Find u’u if u '  3 2 5 .
 More generally, if u '  u1 u2 ... un  , then
u’u =
Length of a vector
 The length of a vector u’ is denoted as ||u||
 Can be written as u  u'u

Special matrices
Special matrices: Identity Matrix
 In scalar algebra, multiplying by 1 changes nothing:

(1)k = k(1) = k
 In matrix algebra, a similar role is played by identity
matrix, I:
IA = AI = A
 How does I look like?
1 0 0 
1 0   
I2    I 3  0 1 0 
 0 1  0 0 1 
Identity Matrix 2
a b
 c d  by I
 Ex: Post-multiply and pre-multiply A 
 e f 
Identity Matrix 3
 Some implications:
1) Inserting or deleting I does not affect the matrix

product:
AIB = (AI)B = AB
2) Multiplying In by itself:
InIn = (In)2 = In
The Null Matrix
 The null matrix (or zero matrix) is a matrix whose
elements are all zero
 Denoted 0
 Properties:
 A+0=0+A=A
 0A = 0
 A0 = 0
 Note: 0 on the right side of = may have a different

dimension in 0A than in A0
Idempotent Matrix
 I and (square) 0 are examples of Idempotent
Matrices
 Def: If A2 = A then matrix A is said to be idempotent.
 Used frequently in econometrics
1 0
 Another example: A   
 0 0 
Idempotent Matrix 2
 Q: Can a non-square matrix be idempotent?

Matrices vs Scalars
 Matrices do not always behave like scalars
1) For scalars, ab = 0 ==> either a=0 or b=0 (or both).

NOT true for matrices.
 2 4 - 2 - 4 
 Ex: A  B 
 1 2   1 2 
Matrices vs Scalars 2
2) For scalars, ab = ac (with a ≠ 0) ==> b = c.
NOT true for matrices.
 2 3 1 1  - 2 1 
 Ex: A  B  C 
 6 9  1 2   3 2 
Rules of matrix
algebra
Commutative, Associative, and
Distributive Laws
In ordinary scalar algebra, addition and multiplication
obey the following laws:
 Commutative law of addition: a+b=b+a
 Commutative law of multiplication: ab = ba
 Associative law of addition: (a+b)+c = a+(b+c)
 Associative law of multiplication: ab(c) = a(bc)
 Distributive law: a(b+c) = ab + ac

001 - Lecture - 3 - Annotated - Matrix Multiplication

Uploaded by

001 - Lecture - 3 - Annotated - Matrix Multiplication

Uploaded by

Matrix multiplication

 Suppose A and B can be multiplied to form a matrix C

 cik is the inner product of

 Ex:  a11 a12   b11 b12   c11 c12 

 Ex: b11 b12 

 Matrix multiplication – examples

 Rules of matrix algebra

 Note: AB and BA are different things!

 Recall: Vectors can be viewed as special types of

 However: If x is a column matrix and a y’ is a row

 Ex 3: Find u’u if u '  3 2 5 .

 More generally, if u '  u1 u2 ... un  , then

 The length of a vector u’ is denoted as ||u||

 Can be written as u  u'u

 In scalar algebra, multiplying by 1 changes nothing:

1) Inserting or deleting I does not affect the matrix

 Note: 0 on the right side of = may have a different

 Def: If A2 = A then matrix A is said to be idempotent.

 Used frequently in econometrics

 Q: Can a non-square matrix be idempotent?

1) For scalars, ab = 0 ==> either a=0 or b=0 (or both).

 Commutative law of addition: a+b=b+a

 Commutative law of multiplication: ab = ba

 Associative law of addition: (a+b)+c = a+(b+c)

 Associative law of multiplication: ab(c) = a(bc)

 Distributive law: a(b+c) = ab + ac

You might also like