Professional Documents
Culture Documents
MA 214 Lecture 9
MA 214 Lecture 9
Notation : x = (x1 , x2 , · · · , xn )T ∈ Rn .
Notation : x = (x1 , x2 , · · · , xn )T ∈ Rn .
Example:
1 Euclidean norm on Rn is denoted
v by ∥ · ∥2 , and is defined by
u n
uX
∥x∥2 = t |xi |2 .
i=1
Notation : x = (x1 , x2 , · · · , xn )T ∈ Rn .
Example:
1 Euclidean norm on Rn is denoted
v by ∥ · ∥2 , and is defined by
u n
uX
∥x∥2 = t |xi |2 .
i=1
2 l∞ norm, which is also called maximum norm, on Rn is denoted
by ∥ · ∥∞ , and is defined by
∥x∥∞ = max { |x1 |, |x2 |, · · · , |xn | } .
Notation : x = (x1 , x2 , · · · , xn )T ∈ Rn .
Example:
1 Euclidean norm on Rn is denoted
v by ∥ · ∥2 , and is defined by
u n
uX
∥x∥2 = t |xi |2 .
i=1
2 l∞ norm, which is also called maximum norm, on Rn is denoted
by ∥ · ∥∞ , and is defined by
∥x∥∞ = max { |x1 |, |x2 |, · · · , |xn | } .
3 l1 norm on Rn is denoted by ∥ · ∥1 , and is defined by
∥x∥1 = |x1 | + |x2 | + · · · + |xn |.
MA 214 - NA Spring 2023-24 3 / 46
Matrix Norms Subordinate to a Vector Norm
X
n
∥A∥∞ = max |aij |.
1≤i≤n
j=1
X
n
∥A∥∞ = max |aij |.
1≤i≤n
j=1
is denoted by ∥A∥2 q
∥A∥2 = max |λi |,
1≤i≤n
is denoted by ∥A∥2 q
∥A∥2 = max |λi |,
1≤i≤n
Theorem
Let A be an invertible n × n matrix.
Theorem
Let A be an invertible n × n matrix. Let x and x̃ be the solutions of the
systems
Ax = b and Ax̃ = b̃,
Theorem
Let A be an invertible n × n matrix. Let x and x̃ be the solutions of the
systems
Ax = b and Ax̃ = b̃,
respectively, where b and b̃ are given vectors.
Theorem
Let A be an invertible n × n matrix. Let x and x̃ be the solutions of the
systems
Ax = b and Ax̃ = b̃,
respectively, where b and b̃ are given vectors. Then
∥x − x̃∥ ∥b − b̃∥
≤ ∥A∥ ∥A−1 ∥
∥x∥ ∥b∥
Definition
Let A be an n × n invertible matrix. Let a matrix norm be given that is
subordinate to a vector norm. Then the condition number of the matrix
A is defined as
κ(A) := ∥A∥ ∥A−1 ∥.
∥x − x̃∥ ∥b − b̃∥
≤ ∥A∥ ∥A−1 ∥
∥x∥ ∥b∥
∥x̃ − x∥ ∥A − Ã∥
≤ ∥A∥ ∥A−1 ∥ .
∥x̃∥ ∥A∥
∥x − x̃∥ ∥b − b̃∥
≤ ∥A∥ ∥A−1 ∥
∥x∥ ∥b∥
∥x̃ − x∥ ∥A − Ã∥
≤ ∥A∥ ∥A−1 ∥ .
∥x̃∥ ∥A∥
It is clear that if the condition number is small, then the relative error in
the solution will also be small whenever the relative error in the right
hand side vector is small.
∥x − x̃∥ ∥b − b̃∥
≤ ∥A∥ ∥A−1 ∥
∥x∥ ∥b∥
∥x̃ − x∥ ∥A − Ã∥
≤ ∥A∥ ∥A−1 ∥ .
∥x̃∥ ∥A∥
It is clear that if the condition number is small, then the relative error in
the solution will also be small whenever the relative error in the right
hand side vector is small.
On the other hand, if the condition number is large, then the relative
error could be very large even though the relative error in the right hand
side vector is small. MA 214 - NA Spring 2023-24 11 / 46
Condition Number of a Matrix (contd.)
Theorem
Let A ∈ Mn (R) be non-singular. Then, for any singular n × n matrix B,
we have
1 ∥A − B∥
≤ .
κ(A) ∥A∥
We have
1 1 1
1
1 1
= = ≤
κ(A) ∥A∥∥A ∥ ∥A∥
−1 ∥A−1 x∥ ∥A∥ ∥A−1 y∥
max
x̸=0 ∥x∥ ∥y∥
Ax = b ⇒ Dx = Cx + b.
Ax = b ⇒ Dx = Cx + b.
Choose (arbitrarily) some specific value for x, say x = x(0) , then the
resulting system
Dx = Cx(0) + b
can be readily solved.
Ax = b ⇒ Dx = Cx + b.
Choose (arbitrarily) some specific value for x, say x = x(0) , then the
resulting system
Dx = Cx(0) + b
can be readily solved.
Let us call the solution of this system as x(1) . That is,
Dx(1) = Cx(0) + b.
x1 + 4x2 + 0.5x3 = 1,
6x1 + x2 + 2x3 = −2,
−x1 + 0.5x2 − 4x3 = 0.
x1 + 4x2 + 0.5x3 = 1,
6x1 + x2 + 2x3 = −2,
−x1 + 0.5x2 − 4x3 = 0.
x1 + 4x2 + 0.5x3 = 1,
6x1 + x2 + 2x3 = −2,
−x1 + 0.5x2 − 4x3 = 0.
Theorem
If the coefficient matrix A is diagonally dominant, then the Jacobi
method
x(k+1) = Bx(k) + c, k = 0, 1, 2, · · · ,
converges.
Theorem
If the coefficient matrix A is diagonally dominant, then the Jacobi
method
x(k+1) = Bx(k) + c, k = 0, 1, 2, · · · ,
converges.
Proof: Recall that the components of the Jacobi iterating sequence x(k)
is given by
1 Xn
= bi − aij xj , i = 1, 2, · · · , n.
(k+1) (k)
xi
aii
j=1,j̸=i
MA 214 - NA Spring 2023-24 28 / 46
Iterative Methods: Jacobi Method (contd.)
This gives X
n
(k+1) aij
|ei | ≤ ∥e(k) ∥∞ .
aii
j=1
j̸=i
This gives X
n
(k+1) aij
|ei | ≤ ∥e(k) ∥∞ .
aii
j=1
j̸=i
Define X
n
aij
µ = max .
1≤i≤n aii
j=1
j̸=i
MA 214 - NA Spring 2023-24 29 / 46
Iterative Methods: Jacobi Method (contd.)
We have
(k+1)
|ei | ≤ µ∥e(k) ∥∞ ,
which is true for all i = 1, 2, · · · , n.
We have
(k+1)
|ei | ≤ µ∥e(k) ∥∞ ,
which is true for all i = 1, 2, · · · , n.
Since this is true for all i = 1, 2, · · · , n, we have
∥e(k+1) ∥∞ ≤ µ∥e(k) ∥∞ .
We have
(k+1)
|ei | ≤ µ∥e(k) ∥∞ ,
which is true for all i = 1, 2, · · · , n.
Since this is true for all i = 1, 2, · · · , n, we have
∥e(k+1) ∥∞ ≤ µ∥e(k) ∥∞ .
We have
(k+1)
|ei | ≤ µ∥e(k) ∥∞ ,
which is true for all i = 1, 2, · · · , n.
Since this is true for all i = 1, 2, · · · , n, we have
∥e(k+1) ∥∞ ≤ µ∥e(k) ∥∞ .
We have
(k+1)
|ei | ≤ µ∥e(k) ∥∞ ,
which is true for all i = 1, 2, · · · , n.
Since this is true for all i = 1, 2, · · · , n, we have
∥e(k+1) ∥∞ ≤ µ∥e(k) ∥∞ .
Let
(0) (0) (0)
x(0) = (x1 , x2 , x3 )T
be an initial guess to the true solution x.
Let
(0) (0) (0)
x(0) = (x1 , x2 , x3 )T
be an initial guess to the true solution x.
The iteration sequence (for k = 0, 1, 2, · · · ) is given by
(k+1) 1 (k) (k)
x1 = (b1 − a12 x2 − a13 x3 )
a11
(k+1) 1 (k) (k)
x2 = (b2 − a21 x1 − a23 x3 )
a22
(k+1) 1 (k) (k)
x3 = (b3 − a31 x1 − a32 x2 ).
a33
MA 214 - NA Spring 2023-24 35 / 46
Iterative Methods: Gauss-Seidel Method (contd.)
Let
(0) (0) (0)
x(0) = (x1 , x2 , x3 )T
be an initial guess to the true solution x.
Define a sequence of iterates (for k = 0, 1, 2, · · · ) by
(k+1) 1 (k) (k)
x1 = (b1 − a12 x2 − a13 x3 )
a11
(k+1) 1 (k+1) (k)
x2 = (b2 − a21 x1 − a23 x3 )
a22
(k+1) 1 (k+1) (k+1)
x3 = (b3 − a31 x1 − a32 x2 ).
a33
MA 214 - NA Spring 2023-24 36 / 46
Iterative Methods: Gauss-Seidel Method (contd.)
Let
(0) (0) (0)
x(0) = (x1 , x2 , x3 )T
be an initial guess to the true solution x.
Define a sequence of iterates (for k = 0, 1, 2, · · · ) by
(k+1) 1 (k) (k)
x1 = (b1 − a12 x2 − a13 x3 )
a11
(k+1) 1 (k+1) (k)
x2 = (b2 − a21 x1 − a23 x3 )
a22
(k+1) 1 (k+1) (k+1)
x3 = (b3 − a31 x1 − a32 x2 ).
a33
This method is called the Gauss-Seidel
MA 214 - NA
Iteration method. Spring 2023-24 36 / 46
Iterative Methods: Gauss-Seidel Method (contd.)
Theorem
If the coefficient matrix is diagonally dominant, then the Gauss-Seidel
method converges.
Define X
i−1
aij X
n
aij
αi = , βi = , i = 1, 2, · · · , n,
aii aii
j=1 j=i+1
with the convention that α1 = βn = 0.
MA 214 - NA Spring 2023-24 38 / 46
Iterative Methods: Gauss-Seidel Method (contd.)
(k+1)
Xi−1
aij (k+1) X aij (k)
n
ei =− e − e , i = 1, 2, · · · , n.
aii j aii j
j=1 j=i+1
Define X
i−1 X
n
aij aij
αi = , βi = , i = 1, 2, · · · , n,
aii aii
j=1 j=i+1
Define X
i−1 X
n
aij aij
αi = , βi = , i = 1, 2, · · · , n,
aii aii
j=1 j=i+1
Define X
i−1 X
n
aij aij
αi = , βi = , i = 1, 2, · · · , n,
aii aii
j=1 j=i+1
(k+1)
⇒ |ei | ≤ αi ∥e(k+1) ∥∞ + βi ∥e(k) ∥∞ , i = 1, 2, · · · , n. (3)
Let l be such that (k+1)
∥e(k+1) ∥∞ = |el |.
Then with i = l in (3),
Observe
X
n
aij
µ = max = max (αi + βi )
1≤i≤n aii 1≤i≤n
j=1
j̸=i
Observe
X
n
aij
µ = max = max (αi + βi )
1≤i≤n aii 1≤i≤n
j=1
j̸=i
Observe
X
n
aij
µ = max = max (αi + βi )
1≤i≤n aii 1≤i≤n
j=1
j̸=i
∥e(k+1) ∥∞ ≤ η∥e(k) ∥∞ .
MA 214 - NA Spring 2023-24 44 / 46
Iterative Methods: Gauss-Seidel Method (contd.)
Define
βi
η = max .
1≤i≤n 1 − αi
Then the above inequality yields
∥e(k+1) ∥∞ ≤ η∥e(k) ∥∞ .
For each i,
βi αi [1 − (αi + βi )] αi
(αi + βi ) − = ≥ [1 − µ] ≥ 0,
1 − αi 1 − αi 1 − αi
⇒ η ≤ µ < 1.
Thus, when the coefficient matrix A is diagonally dominant,
Gauss-Seidel method converges. MA 214 - NA Spring 2023-24 46 / 46