Math281-Linear Algebra

Chapter 7: Matrices
Dr. Mohamad El Dandachy
[email protected]

Eng.Tarek Halawani
[email protected]

1
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
2
Definition
▪ A matrix is a rectangular arrangement of numbers (or functions) in rows
and columns
𝑎11 𝑎12 … 𝑎1𝑛
𝑎21 𝑎22 … 𝑎2𝑛
𝐀 = 𝐴 = 𝑎𝑖 𝑗 = ⋮ ⋮ ⋱ ⋮
𝑚 ×𝑛
𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎 𝑚𝑛
▪ Rows run horizontally and columns run vertically.
▪ The above matrix has m rows and n columns.

▪ Size of the matrix: 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑜𝑤𝑠 × 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑙𝑢𝑚𝑛𝑠

▪ The above matrix has is of a size 𝑚 × 𝑛 (m by n)

▪ Entries or Elements of the matrix: The number in the matrix

▪ The element 𝒂𝒊𝒋 is the element in the ith row and jth column
3
Special Type of matrices
▪ A 𝟏 × 𝟏 matrix is a constant or a single function or scalar matrix

▪ A 𝟏 × 𝒏 matrix: 𝐀 = 𝑎1 𝑎2 … 𝑎 𝑛 is a row matrix (or row vector)

𝑏1
𝑏
▪ A 𝒎 × 𝟏 matrix: 𝐁 = 2 is a column matrix (or column vector)
⋮
𝑏𝑚

▪ A zero matrix: is a matrix (of any size) whose all elements are zero
𝑐11 𝑐12 … 𝑐1𝑚
𝑐21 𝑐22 … 𝑐2𝑚
▪ A 𝒎 × 𝒎 matrix: (m=n) 𝐂 = ⋮ ⋮ ⋱ ⋮ is a square matrix
𝑐𝑚1 𝑐𝑚2 ⋯ 𝑐 𝑚𝑚
▪ In a square matrix the elements {𝑎𝑖𝑖} = {𝑎11, 𝑎22,…, 𝑎 𝑚𝑚 } are called the main or principal diagonal of the
matrix
▪ Square matrix are important when solving sets of linear equations.
4
Special Type of square matrices
▪ Symmetric matrix: is a square matrix with 𝒂𝒊𝒋 = 𝒂𝒋𝒊 for all i and j
5 1 2
ex: 𝐀 = 1 6 −7
2 −7 8

▪ Skew Symmetric matrix: is a square matrix with 𝒂𝒊𝒋 = −𝒂𝒋𝒊 for all i and j
0 1 −2
ex: 𝐁 = −1 0 7
2 −7 0

▪ diagonal matrix: is a square matrix with 𝒂𝒊𝒊 ≠ 𝟎 and 𝒂𝒊𝒋 = 𝟎 for all 𝑖 ≠ 𝑗
5 0 0
ex: 𝐂 = 0 6 0
0 0 8
▪ Identity matrix: is a square matrix with 𝒂𝒊𝒊 = 𝟏 and 𝒂𝒊𝒋 = 𝟎 for all 𝑖 ≠ 𝑗 denoted by 𝑰
1 0 0
1 0
▪ ex: 𝐈𝟑×𝟑 = 0 1 0 , 𝐈𝟐×𝟐 =
0 1
0 0 1
▪ The multiplication of any 𝐀 matrix by 𝐈 gives the matrix 𝐀 → 𝑨𝐈 = 𝐀 5
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
6
Arithmetic operation on matrices
▪ Equality: Two matrices 𝐀 = 𝑎 𝑖𝑗 and 𝐁 = 𝑏𝑖𝑗 are equal if:
𝑚×𝑛 𝑟×𝑐
▪ They have the same size (𝑚 = 𝑟 and 𝑛 = 𝑐)
▪ 𝑎 𝑖𝑗 = 𝑏𝑖𝑗 for all i and j

▪ Addition of two matrices:

▪ To add matrices, they must be of the same order
▪ 𝐀 + 𝐁 = 𝑎 𝑖𝑗 +𝑏𝑖𝑗 (add their corresponding entries.)
𝑚 ×𝑛
2 −1 3 4 7 −8 2+4 −1 + 7 3 + (−8) 6 6 −5
▪ Ex: 0 4 6 + 9 3 5 = 0+9 4+3 6 + 5 = 9 7 11
−6 10 −5 1 −1 2 −6 + 1 10 + (−1) −5 + 2 −5 9 −3

▪ Subtraction of two matrices:

▪ To subtract matrices, they must be of the same order
▪ 𝐀 − 𝐁 = 𝑎 𝑖𝑗 −𝑏𝑖𝑗 (subtract their corresponding entries.)
𝑚 ×𝑛
2 −1 3 4 7 −8 2−4 −1 − 7 3 − (−8) −2 −8 11
▪ Ex: 0 4 6 − 9 3 5 = 0−9 4−3 6 − 5 = −9 1 1
−6 10 −5 1 −1 2 −6 − 1 10 − (−1) −5 − 2 −7 11 −7
7
Arithmetic operation on matrices (continue)
▪ Multiplication by a scalar:
▪ 𝑘 ∙ 𝐀 = 𝑘 𝑎𝑖𝑗 = 𝑘𝑎𝑖𝑗
𝑚×𝑛 𝑚×𝑛
▪ multiply each entry of A by the real number k
▪ 𝑘 ∙ 𝐀 = 𝐀 ∙ 𝑘 (scalar multiplication follow the commutative law)

· 2 −3 −5 × 2 −5 × −3 −10 15
ex 1: −5 = =
4 −1 −5 × 4 −5 × −1 −20 5

2 −3 ∙ 2 × −5 −3 × −5 −10 15
ex 2: −5 = =
4 −1 4 × −5 −1 × −5 −20 5
8
 Arithmetic operation on matrices (continue)
▪ Matrix Multiplication:
▪ Let 𝐀 = 𝑎 𝑖𝑗 and 𝐁 = 𝑏𝑖𝑗
𝑚 ×𝑛 𝑝 ×𝑞
▪ 𝐂 = 𝐀 ∙ 𝐁 = 𝐀𝐁 is only defined if 𝒏 = 𝒑 and 𝐂 will be a 𝒎 × 𝒒 matrix
▪ You can only multiply if the number of columns in the 1st matrix is equal to the number of rows in the 2nd matrix.
▪ ex:
▪ 𝐀𝟐×𝟐 ∙ 𝑩𝟑×𝟐is not defined
▪ 𝐀𝟐×𝟑 ∙ 𝑩𝟑×𝟓 ∙ 𝑪𝟓×𝟒 gives a matrix of size 2 × 4
▪ The entries 𝑐 𝑖𝑗 = ∑ 𝒏𝑘 =1 𝑎 𝑖𝑘 𝑏 𝑘𝑗
▪ The matrix multiplication is not commutative: 𝐀 ∙ 𝐁 ≠ 𝐁 ∙ 𝐀
▪ Matrix multiplication is associative: (𝐀 ∙ 𝐁) ∙ 𝐂 = 𝐀 ∙ (𝐁 ∙ 𝐂)
ex: 𝟐 −𝟏 ∙ · 𝟑 −𝟗 𝟐
=?
𝟏 −𝟐𝟓 𝟏𝟎
𝟑𝟑 𝟒𝟒 𝟓𝟓 𝟕𝟕 −𝟔
−𝟔 𝟐𝟗 𝟏 −𝟏𝟖

MATH281 - BAU - FALL 2018/2019 9

Arithmetic operations on matrices (continue)
▪ Each of the following statements is valid for any scalars 𝛼 and 𝛽 and for any
matrices A, B, and C for which the indicated operations are defined.
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
Transpose of a Matrix
▪ The transpose of a matrix 𝐀 = 𝑎𝑖𝑗 is 𝐀𝑻 = 𝑎𝑗𝑖
𝑚×𝑛 𝑛 ×𝑚

6 6 −5 6 9 −4
▪ ex 1: A = 9 7 11 → AT = 6 7 9
−4 9 −3 −5 11 −3
2 −3
2 4 7
▪ ex 2: 𝐵 = 4 −1 → 𝐵𝑇 =
−3 −1 6
7 6

▪ Let 𝐀 be a square matrix :

▪ If 𝐀𝐓 = 𝐀 → 𝐀 is symmetric matrix
▪ If 𝐀𝐓 = −𝐀 → 𝐀 is skew-symmetric matrix
▪ Any square matrix 𝐀 could be decomposed into 2 matrices one of them is symmetric and the
other is skew-symmetric:
𝐀= 𝐂 + 𝐃
𝟏
▪ 𝐂 = 𝐀 + 𝐀𝑻 is a symmetric matrix
𝟐
𝟏
▪ 𝐃 = 𝐀 − 𝐀𝑻 is skew-symmetric matrix
𝟐
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
Row echelon form

MATH281 - BAU - FALL 2018/2019 14

Reduce a matrix to a Row echelon form
▪ Gauss elimination: To transform the matrix to a row echelon form
▪ Step 1: the first non zero entry in row 1 should be 1
▪ Step 2: the first entries in below rows should be zeros
▪ Step 3: the second entry in row 2 should be 1
▪ Step 4: the second entries in below rows should be zeros
▪ And so on ….
 Reduce a matrix to a Row echelon form:
Gauss elimination
Example: Find the row echelon form of A = 4 5 −6 1
7 −3 0 8

▪ Step 1: the first non zero entry in row 1 should be 1→ divide row 1 by 4
𝑅1
5 3 1
4 5 −6 1 4 1 −
7 −3 0 8 4 2 4
7 −3 0 8
▪ Step 2: the first entries in below rows should be zeros → subtract 7 Row1 from Row 2
5 3 1
5 3 1
1 𝑅2−7𝑅1 1 4
−
2 4
−
4 2 4 47 21 25
7 −3 0 8 0 −
4 2 4
▪ Step 3: the second entry in row 2 should be 1
5 3 1 47 𝟓 𝟑 𝟏
1 − 𝑅 /− 𝟏 −
4 2 4 2 4 𝟒 𝟐 𝟒
𝑺𝑻𝑶𝑷
47 21 25 𝟒𝟐 𝟐𝟓
0 𝟎 𝟏 −
4 2 4 𝟒𝟕 𝟒𝟕
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
 Rank of a matrix
▪ The rank of 𝑚 × 𝑛 matrix 𝐀 is denoted by 𝑟𝑎𝑛𝑘 𝐀

▪ 𝑟𝑎𝑛𝑘 𝐀 ≤ min(𝑚, 𝑛)

▪ 𝑟𝑎𝑛𝑘 𝐀 = 𝑟𝑎𝑛𝑘(𝐀𝐓)

▪ 𝑟𝑎𝑛𝑘 𝐀 can be found by:

▪ Method 1: the maximum number of linearly independent column (or rows) vectors in A.
𝑎11 𝑎12 … 𝑎1𝑛 𝑎1𝑗
𝑎21 𝑎22 … 𝑎2𝑛 𝑎2𝑗
▪ Any matrix A = ⋮ ⋮ ⋱ ⋮ consists of n vectors 𝑣 = ⋮
𝑎𝑚1 𝑎𝑚2 ⋯ 𝑎 𝑚𝑛 𝑎 𝑚𝑗
▪ These vectors are linearly independent if : 𝛽1𝑣1 + 𝛽2𝑣2 + ⋯ + 𝛽𝑛𝑣𝑛 = 0 only and only for 𝛽1 = 𝛽2 = ⋯ = 𝛽𝑛 = 0

▪ Method 2: the number leading coefficients in the echelon form (see slides 16 and 17)

▪ Method 3: Equal to the dimension of the largest square sub-matrix of A that has a non-zero
determinant. 18
 Rank of a matrix: Example
▪ 𝑟𝑎𝑛𝑘 C ≤ min(𝑚, 𝑛) ≤ 4
▪ 𝑟𝑎𝑛𝑘 C can be found by:
▪ Method 2: the number leading coefficients in the echelon form

19
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
20
The inverse of a square matrix
▪ An 𝑛 × 𝑛 matrix 𝐀 is said to be nonsingular or invertible if there
exists a matrix 𝐵 such that 𝐀𝐁 = 𝐁𝐀 = 𝐈

▪ The matrix 𝐁 is said to be a multiplicative inverse of 𝐀 and is denoted

by 𝐀−𝟏 (𝐀−𝟏 is not a reciprocal)

▪ An 𝑛 × 𝑛 matrix 𝐀 is said to be singular if it does not have a

multiplicative inverse ( 𝐀−𝟏 doesn’t exist or 𝐀−𝟏 = [0]).

▪ Only square matrices may have multiplicative inverses.

21
The inverse of a square matrix
▪ The inverse of 𝐀 is obtained by:
𝟏
𝐀−𝟏 = 𝒂𝒅𝒋(𝐀)
𝒅𝒆𝒕(𝐀)
▪ 𝒅𝒆𝒕(𝑨) ≠ 𝟎 is the determinant of 𝐀 (see chapter 6)
▪ if 𝒅𝒆𝒕 𝑨 = 𝟎 the inverse doesn’t exist and matrix is singular

▪ 𝒂𝒅𝒋(𝐀) is the adjoint matrix of 𝐀 and it is defined as the transpose of the matrix
of the signed minor determinant corresponding to the elements of 𝐀
1+𝑛 𝑀 𝑻
+𝑀11 −𝑀12 ⋯ −1 1𝑛
−𝑀21 +𝑀22 ⋯ −1 2+𝑛𝑀
2𝑛
𝒂𝒅𝒋 𝐀 = ⋮ ⋮ ⋱ ⋮
−1 𝑛+1𝑀𝑛1 −1 𝑛+2𝑀𝑛2 ⋯ −1 𝑛+𝑛 𝑀
𝑛𝑛

▪ 𝐶𝑖𝑗 = −1 𝑖+𝑗 𝑀𝑖𝑗 is the signed minor determinant (see chapter 6) and is called cofactor
22
 The inverse of a square matrix
4 -3
Example 1: Find the inverse of : 𝐀 = and check your answer
-𝟐 5
𝟏
𝐀−𝟏 = 𝒂𝒅𝒋(𝐀)
𝒅𝒆𝒕(𝐀)

23
 The inverse of a square matrix
Example 2: Find the inverse of : and check your answer

24
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
25
Solving of linear system of equations
using inverse matrix
𝑎11 𝑥 + 𝑎12 𝑦 + 𝑎13 𝑧 = 𝑏1 ▪ Example 1:
▪ A linear system: 𝑎21 𝑥 + 𝑎22 𝑦 + 𝑎23𝑧 = 𝑏2
𝑎31 𝑥 + 𝑎32 𝑦 + 𝑎33𝑧 = 𝑏3 ▪ Use the inverse of
matrix to solve:

can be written in matrix form: 𝐴 ∙ 𝑋 = 𝐵 4𝑥 − 3𝑦 = 6

a.
𝑎11 𝑎12 𝑎13 𝑥 𝑏1 −2𝑥 + 5𝑦 = 4
—With: 𝐴 = 𝑎21 𝑎22 𝑎23 , 𝑋 = 𝑦 and 𝐵 = 𝑏2
𝑎31 𝑎32 𝑎33 𝑧 𝑏3
𝑥 + 2 𝑦 + 3 𝑧 = −4
b. 4 𝑥 + 5𝑧 = −11
▪ If 𝒅𝒆𝒕 𝑨 ≠ 𝟎 the solution of this system can be obtained by: 6 𝑥 − 7 𝑦 + 𝑧 = −11

𝑿 = 𝑨−𝟏 ∙ 𝑩

26
 Solving of linear system of equations
using inverse matrix
𝑥 + 2 𝑦 + 3 𝑧 = −4
b. 4 𝑥 + 5𝑧 = −11
6 𝑥 − 7 𝑦 + 𝑧 = −11

27
Solving of linear system of equations
using Gaussian Elimination
𝑎11 𝑥 + 𝑎12 𝑦 + 𝑎13 𝑧 = 𝑏1
▪ A linear system: 𝑎21 𝑥 + 𝑎22 𝑦 + 𝑎23𝑧 = 𝑏2
𝑎31 𝑥 + 𝑎32 𝑦 + 𝑎33𝑧 = 𝑏3
𝒂𝟏𝟏 𝒂𝟏𝟐 𝒂𝟏𝟑 𝒃𝟏
can be represented by the augmented matrix: 𝒂𝟐𝟏 𝒂𝟐𝟐 𝒂𝟐𝟑 𝒃𝟐
𝒂𝟑𝟏 𝒂𝟑𝟐 𝒂𝟑𝟑 𝒃𝟑

▪ Transform this augmented matrix to row echelon form matrix using the Elementary row operations (see slides
14, 15 and 16)
1 𝑑12 𝑑13 𝑐1
0 1 𝑑23 𝑐2
0 0 1 𝑐3
▪ This form gives the following linear system:
𝑥 + 𝑑12 𝑦 + 𝑑13 𝑧 = 𝑐1
𝑦 + 𝑑23𝑧 = 𝑐2
𝑧 = 𝑐3
▪ By back substitution we obtain the solution:
𝑧 = 𝑐3
𝑦 + 𝑑23 𝑐3 = 𝑐2 → 𝑦 = 𝑐2 − 𝑑23𝑐3
𝑥 + 𝑑12 𝑐2 − 𝑑23𝑐3 + 𝑑13 𝑐3 = 𝑐1 → 𝑥 = 𝑐1 − 𝑑12 𝑐2 − 𝑑23𝑐3 − 𝑑13 𝑐3
28
Solving of linear system of equations
using Gauss-Jordan Elimination
𝑎11 𝑥 + 𝑎12 𝑦 + 𝑎13 𝑧 = 𝑏1
▪ A linear system: 𝑎21 𝑥 + 𝑎22 𝑦 + 𝑎23𝑧 = 𝑏2
𝑎31 𝑥 + 𝑎32 𝑦 + 𝑎33𝑧 = 𝑏3
𝑎11 𝑎12 𝑎13 𝑏1
can be represented by the augmented matrix: 𝑎21 𝑎22 𝑎23 𝑏2
𝑎31 𝑎32 𝑎33 𝑏3

▪ Transform this augmented matrix to reduced-row echelon form matrix using the
Elementary row operations (see slides 14, 15 and 16)
𝟏 𝟎 𝟎 𝒄𝟏
Example on
𝟎 𝟏 𝟎 𝐜𝟐
reduced-row echelon form
𝟎 𝟎 𝟏 𝒄𝟑
▪ This form gives the solution of the linear system:
𝑥 = 𝑐1
𝑦 = 𝑐2
𝑧 = 𝑐3
▪ No need to back substitution
29
Example 2:
2𝑥 + 6 𝑦 + 𝑧 = 7
• Find the solution of: 𝑥 + 2𝑦 − 𝑧 = −1 using:
5 𝑥 + 7 𝑦 − 4𝑧 = 9
a. Gaussian Elimination (row echlon form)
b. Gauss-Jordan elimination (reduced row echlon form)

30
31
32
9

33
34
Modify the matrix from row echlon form to row-reduced echlon form

35
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
36
The eigenvalues and eigenvectors problem
▪ Let 𝐀 be 𝑛 × 𝑛 matrix, a number 𝜆 is the eigenvalue of 𝐀 if there exists a non-zero solution vector 𝐗 = 𝑲 of
the linear system
𝐀∙ 𝑿=𝝀 𝑿
Or 𝐀 − 𝝀 𝑰 𝑿 = 𝟎

▪ The solution vector 𝐗 = 𝑲 is the eigenvector corresponding to the eigenvalue 𝝀

▪ As we are searching for non-Trivial solutions 𝐗 ≠ 𝟎, the system has this solution if:
𝐝𝐞𝐭 𝐀 − 𝝀 𝑰 = 𝟎 (characteristic equation of A)

▪ The eigenvalues are the roots of the characteristic equation of A:

▪ Application of eigenvalue:
▪ Differential equations
▪ Communication systems: the theoretical limit of communication channel capacity or steady state of
Markov chains
▪ Wherever there are vibrations, there are eigenvalues, the natural frequencies of the vibrations.
▪ When engineers design structures, they are concerned with the frequencies of vibration of the
structure. (Designing bridges for
example) 37
Find the eigenvalues and eigenvectors of a
matrix
▪ If the eigenvector 𝑿 = 𝑲 is given:
▪ We find 𝐀 ∙ 𝑲
▪ It should be in the form 𝝀𝑲 → 𝝀 is an eigenvalue

▪ If the eigenvalue 𝝀 is given:

▪ Resolve A − 𝜆 𝐼 𝑋 = 0 (using Gauss-Jordan for example)
▪ the solution 𝑋 is the eigenvector corresponding to the eigenvalue 𝜆

▪ General case:
▪ Find First the eigenvalues: Find the roots of det A − 𝜆 𝐼 = 0
▪ For each eigenvalue 𝝀 : Find the solution of A − 𝜆 𝐼 𝑋 = 0 (using Gauss-Jordan for
example), The solution of the X = 𝑲 is the eigenvector corresponding to 𝜆
38
Example 1:

39
Example 2:

40
Example 2:

41
0
42
0
43
44
45
Content
▪ General properties of matrices
▪ Arithmetic operation on matrices
▪ Transpose of a Matrix
▪ Row echelon form
▪ Rank of a matrix
▪ The inverse of a square matrix
▪ Solving of linear system of equations
▪ By using inverse matrix
▪ By Gaussian Elimination
▪ The eigenvalues and eigenvectors problem
▪ Matrix partitioning (block decomposition)
46
 Matrix partitioning
▪ A matrix 𝐀 can be partitioned into smaller matrices called block or submatrix

▪ The partition is indicated by horizontal lines between the rows and vertical lines between the
columns.
row vector
a11 a12 a13 a14  column vector
a11 a14  r1 
A = a 21 a 22 a 23 a 24  = c1 c 2 c3 c 4 
a12 a13
A = a 21 a22 a23 a24  = r2  
   a31 a32 a33 a 34 
 a31 a32 a33 a34  r3 
submatrix
a11 a12 a13 a14 
  A11 A12 
A = a21 a22 a23 a24 = 
   A21 A22 
a31 a32 a33 a34 
47
Block multiplication
Let 𝐀 be an 𝑚 × 𝒏 matrix and 𝐁 an 𝒏 × 𝑝 matrix

▪ If 𝐁 has been partitioned into columns [𝐛𝟏 𝐛𝟐 … 𝐛𝐫]

▪ 𝐀 ∙ 𝐁 = [𝐀 ∙ 𝐛𝟏 𝐀 ∙ 𝐛𝟐 … 𝐀 ∙ 𝐛𝐫]
𝐚𝟏
𝐚𝟐
▪ If 𝐀 has has been partitioned into rows
..
𝐚𝟏 ∙ 𝐁 𝐚𝐦
𝐚𝟐 ∙ 𝐁
▪ 𝐀∙ 𝐁=
..
𝐚𝐦 ∙ 𝐁

48
Block multiplication
Let 𝐀 be an 𝑚 × 𝒏 matrix and 𝐁 an 𝒏 × 𝑝 matrix
▪ In general we may decompose 𝐀 and 𝐁

With
▪𝐀𝟏𝟏 is a 𝑟 × 𝑠 matrix ▪𝐁𝟏𝟏 is a s × 𝑞 matrix
▪𝐀𝟏𝟐 is a 𝑟 × (𝑛 − 𝑠) matrix ▪𝐁𝟏𝟐 is a s × (𝑝 − 𝑞) matrix
▪𝐀𝟐𝟏 is a (𝑚 − 𝑟) × 𝑠 matrix ▪ 𝐁𝟐𝟏 is a (𝑛 − 𝑠) × 𝑞 matrix
▪𝐀𝟐𝟐 is a(𝑚 − 𝑟) × (𝑛 − 𝑠) matrix ▪𝐁𝟐𝟐 is a(𝑛 − 𝑠) × (𝑝 − 𝑞) matrix

𝐀𝟏𝟏𝐁𝟏𝟏 + 𝐀𝟏𝟐𝐁𝟐𝟏 𝐀𝟏𝟏𝐁𝟏𝟐 + 𝐀𝟏𝟐𝐁𝟐𝟐

▪ Thus 𝐀 ∙ 𝐁 =
𝐀𝟐𝟏𝐁𝟏𝟏 + 𝐀𝟐𝟐𝐁𝟐1 𝐀𝟐1 𝐁1𝟐+𝐀𝟐𝟐𝐁𝟐𝟐 49