Convergence Criteria

1
Iterative Solution Methods






Starts with an initial approximation for the
solution vector (x0)
At each iteration updates the x vector by using
the sytem Ax=b
During the iterations A, matrix is not changed
so sparcity is preserved



Each iteration involves a matrix-vector product



If A is sparse this product is efficiently done
2
Iterative solution procedure








Write the system Ax=b in an equivalent form
x=Ex+f (like x=g(x) for fixed-point iteration)
Starting with x0, generate a sequence of
approximations {xk} iteratively by
x k+1 =Ex k +f
Representation of E and f depends on the type
of the method used
But for every method E and f are obtained from
A and b, but in a different way
3
Convergence








As k→∞, the sequence {xk} converges to the
solution vector under some conditions on E
matrix
This imposes different conditions on A matrix
for different methods
For the same A matrix, one method may
converge while the other may diverge
Therefore for each method the relation
between A and E should be found to decide on
the convergence
4
Different Iterative methods




Jacobi Iteration
Gauss-Seidel Iteration
Successive Over Relaxation (S.O.R)




SOR is a method used to accelerate the
convergence
Gauss-Seidel Iteration is a special case of SOR
method

5
Jacobi iteration
a11 x1 + a12 x2 +  + a1n xn = b1
a21 x1 + a22 x2 +  + a2 n xn = b2

an1 x1 + an 2 x2 +  + ann xn = bn

 x10 
 0
0
 x2 
x =

 0
 xn 
 

1
0
0
1 
(b1 − a12 x2 −  − a1n xn )
k +1
xi = bi −
a11
aii 
1
0
0
x1 =
(b2 − a21 x10 − a23 x3 −  − a2 n xn )
2
a22
1
0
0
x1 =
(bn − an1 x10 − an 2 x2 −  − ann −1 xn −1 )
n
ann
1
x1 =


∑ aij x − j∑1aij x 
j =1
=i +

i −1

k
j

n

k
j

6
xk+1=Exk+f iteration for Jacobi method
A can be written as A=L+D+U (not decomposition)
0 0  a11 0
0  0 a12 a13 
 a11 a12 a13   0
a
a22 a23  =  a21 0 0 +  0 a22 0  + 0 0 a23 
 21
 
 
 

 a31 a32 a33   a31 a32 0  0
0 a33  0 0
0

 
 
 


Ax=b ⇒ (L+D+U)x=b
1
k +1
xi =
aii
Dxk+1

i −1
n


k
k
bi − ∑ aij x j − ∑ aij x j 
j =1
j = i +1





Lx k



Uxk

Dxk+1 =-(L+U)xk+b
xk+1=-D-1(L+U)xk+D-1b
E=-D-1(L+U)
f=D-1b
7
Gauss-Seidel (GS) iteration
Use the latest
update

a11 x1 + a12 x2 +  + a1n xn = b1
a21 x1 + a22 x2 +  + a2 n xn = b2

an1 x1 + an 2 x2 +  + ann xn = bn

1
0
0
1 
(b1 − a12 x2 −  − a1n xn )
k +1
xi = bi −
a11
aii 
1
1
1
0
0
x2 =
(b2 − a21 x1 − a23 x3 −  − a2 n xn )
a22
1
1
x1 =
(bn − an1 x1 − an 2 x1 −  − ann −1 x1 −1 )
n
2
n
ann
1
x1 =

 x10 
 0
0
 x2 
x =

 0
 xn 
 
i −1

∑a x
j =1

ij

k +1
j


− ∑ aij x 
j = i +1

n

k
j

8
x(k+1)=Ex(k)+f iteration for Gauss-Seidel
Ax=b ⇒ (L+D+U)x=b
k +1
i

x

Dx

1
=
aii

k+1

i −1
n


k +1
k
bi − ∑ aij x j − ∑ aij x j 
j =1
j = i +1





Lx k +1



(D+L)xk+1 =-Uxk+b

Uxk

xk+1=-(D+L)-1Uxk+(D+L)-1b
E=-(D+L)-1U
f=-(D+L)-1b
9
Comparison






Gauss-Seidel iteration converges more rapidly
than the Jacobi iteration since it uses the latest
updates
But there are some cases that Jacobi iteration
does converge but Gauss-Seidel does not
To accelerate the Gauss-Seidel method even
further, successive over relaxation method can
be used
10
Successive Over Relaxation Method


GS iteration can be also written as follows
k +1
i

x

1
=x +
aii
k
i

i −1
n


k +1
k
bi − ∑ aij x j − ∑ aij x j 
j =1
j =i



xik +1 = xik + δ ik

Correction term
ωδ i2

xi3
xi2
xi1
0
i

x

δ

ωδ i1

2
i

δ i1

δ i0

Multiply with

ω >1

Faster
convergence

ωδ i0
11
SOR
xik +1 = xik + ωδ ik
i −1
n


k +1
k
x
bi − ∑ aij x j − ∑ aij x j 
j =1
j =i


i −1
n

1 
k +1
k
k +1
k
xi = (1 − ω ) xi + ω bi − ∑ aij x j − ∑ aij x j 
aii 
j =1
j =i +1

k +1
i

1
= x +ω
aii
k
i

1<ω<2 over relaxation (faster convergence)
0<ω<1 under relaxation (slower convergence)
There is an optimum value for ω
Find it by trial and error (usually around 1.6)
12
x(k+1)=Ex(k)+f iteration for SOR
1
k +1
k
xi = (1 − ω ) xi + ω
aii

i −1
n


k +1
k
bi − ∑ aij x j − ∑ aij x j 
j =1
j = i +1



Dxk+1=(1-ω)Dxk+ωb-ωLxk+1-ωUxk
(D+ ωL)xk+1=[(1-ω)D-ωU]xk+ωb
E=(D+ ωL)-1[(1-ω)D-ωU]
f= ω(D+ ωL)-1b
13
The Conjugate Gradient Method
d 0 = r0 = b − Ax0
T

ri ri
αi = T
d i Ad i
xi +1 = xi + α i Ad i

• Converges if A is a
symmetric positive
definite matrix
• Convergence is
faster

T
i +1 i +1
T
i i

r r
βi +1 =
r r

d i +1 = ri +1 + βi +1d i

14
Convergence of Iterative Methods
ˆ
Define the solution vector as x
k
Define an error vector as e

ˆ
x =e +x
k

k

Substitute this into x

k +1

= Ex + f
k

ˆ
ˆ
ˆ
e k +1 + x = E (e k + x) + f = Ex + f + Ee k
e k +1 = Ee k = EEe k −1 = EEEe k − 2 = E ( k +1) e 0
15
Convergence of Iterative Methods
iteration

e

k +1

= E

( k +1) 0

e ≤ E

( k +1)

e

0

power

The iterative method will converge for any initial
iteration vector if the following condition is satisfied
Convergence condition

Lim e k +1 → 0 if
k →∞

Lim E ( k +1) → 0
k →∞

16
Norm of a vector
A vector norm should satisfy these conditions

x ≥ 0 for every nonzero vector x
x = 0 iff x is a zero vector
αx = α x

for scalar α

x+ y ≤ x + y
Vector norms can be defined in different forms as
long as the norm definition satisfies these conditions
17
Commonly used vector norms
Sum norm or ℓ1 norm

x 1 = x1 + x2 +  + xn
Euclidean norm or ℓ2 norm
2
2
x 2 = x12 + x2 +  + xn

Maximum norm or ℓ∞ norm

x

∞

= max i xi
18
Norm of a matrix
A matrix norm should satisfy these conditions

A ≥0
A = 0 iff A is a zero matrix
for scalar α

αA = α A

A+ B ≤ A + B
Important identitiy

Ax ≤ A x

x is a vector
19
Commonly used matrix norms
Maximum column-sum norm or ℓ1 norm
m

A 1 = max ∑ aij
1≤ j ≤ n

i =1

Spectral norm or ℓ2 norm

A 2 = maximum eigenvalue of AT A
Maximum row-sum norm or ℓ∞ norm
n

A ∞ = max ∑ aij
1≤i ≤ m

j =1

20
Example


Compute the ℓ1 and ℓ∞ norms of the matrix

3 9 5
7 2 4 


6 8 1 


16

19

17 = A ∞
13
15

10

= A1
21
Convergence condition
lim e
k →∞

k +1

→ 0 if

lim E
k →∞

( k +1)

→0

Express E in terms of modal matrix P and Λ
Λ:Diagonal matrix with eigenvalues of E on the diagonal

E = PΛP

−1

E ( k +1) = PΛP −1 PΛP −1  PΛP −1
E ( k +1) = PΛ( k +1) P −1

k
λ1 +1



λk +1
2

Λk +1 = 





λk +1 

n 


lim E ( k +1) → 0 ⇒ lim PΛ( k +1) P −1 → 0 ⇒ lim Λ( k +1) → 0
k →∞

k →∞

k →∞

⇒ lim λki +1 → 0 ⇒ λ i < 1 for i = 1,2 ,...,n
k →∞

22
Sufficient condition for convergence
If the magnitude of all eigenvalues of iteration matrix
E is less than 1 than the iteration is convergent
It is easier to compute the norm of a matrix than to
compute its eigenvalues
Ex = λx

Ex = λ x 

⇒ λ x ≤ E x ⇒ λ ≤ E
Ex ≤ E x 


E < 1 is a sufficient condition for convergence
23
Convergence of Jacobi iteration
E=-D-1(L+U)

 0

 − a21
 a22
E= 



 an1
− a
 nn

a12
−
a11
0



a23
−
a22













ann −1
−
ann

a1n 
−
a11 

a2 n 
−
a22 
 
an −1n 
−

an −1n −1 

0 

24
Convergence of Jacobi iteration
Evaluate the infinity(maximum row sum) norm of E
E

n

∞

<1⇒ ∑
j =1
i≠ j

aij

< 1 for i = 1,2,..., n

aii
n

⇒ aii > ∑ aij
j =1
i≠ j

Diagonally dominant matrix

If A is a diagonally dominant matrix, then Jacobi
iteration converges for any initial vector
25
Stopping Criteria


Ax=b



At any iteration k, the residual term is
rk=b-Axk



Check the norm of the residual term
||b-Axk||



If it is less than a threshold value stop

26
Example 1 (Jacobi Iteration)
 4 − 1 1  x1   7 
 4 − 8 1  x  = − 21

 2  

− 2 1 5  x3   15 

  


0 
x 0 = 0 
 
0 
 

b − Ax 0

2

= 26.7395

Diagonally dominant matrix
0
0
7 + x2 − x3
7
x =
= = 1.75
4
4
0
21 + 4 x10 + x3
21
x1 =
= = 2.625
2
8
8
0
15 + 2 x10 − x2
15
1
x3 =
= = 3.0
5
5
1
1

b − Ax1 = 10.0452
2

27
Example 1 continued...
1
7 + x1 − x3
7 + 2.625 − 3
2
x =
=
= 1.65625
4
4
1
1
21 + 4 x1 + x3 21 + 4 × 1.75 + 3
2
x2 =
=
= 3.875
8
8
1
15 + 2 x1 − x1 15 + 2 × 1.75 − 2.625
2
2
=
= 4.225
x3 =
5
5
2
1

7 + 3.875 − 4.225
= 1.6625
4
21 + 4 ×1.65625 + 4.225
3
x2 =
= 3.98125
8
15 + 2 × 1.65625 − 3.875
3
x3 =
= 2.8875
5

b − Ax 2

2

= 6.7413

x13 =

b − Ax 2

2

= 1.9534

Matrix is diagonally dominant, Jacobi iterations are converging
28
Example 2
− 2 1 5  x1   15 
 4 − 8 1  x  = − 21

 2  

 4 − 1 1  x3   7 

  


0 
x 0 = 0 
 
0 
 

b − Ax 0

2

= 26.7395

The matrix is not diagonally dominant
0
0
− 15 + x2 + 5 x3 − 15
x =
=
= −7.5
2
2
0
21 + 4 x10 + x3
21
1
x2 =
= = 2.625
8
8
1
0
x3 = 7 − 4 x10 + x2
= 7.0
1
1

b − Ax1 = 54.8546
2

29
Example 2 continued...
− 15 + 2.625 + 5 × 7
= 11.3125
2
21 − 4 × 7.5 + 7
x1 =
= −0.25
2
8
1
x3 = 7 + 4 × 7.5 + 2.625 = 39.625
1
x1 =

b − Ax 2

2

= 208.3761

The residual term is increasing at each iteration,
so the iterations are diverging.
Note that the matrix is not diagonally dominant
30
Convergence of Gauss-Seidel
iteration






GS iteration converges for any initial vector if A
is a diagonally dominant matrix
GS iteration converges for any initial vector if A
is a symmetric and positive definite matrix
Matrix A is positive definite if
xTAx>0 for every nonzero x vector
31
Positive Definite Matrices






A matrix is positive definite if all its eigenvalues
are positive
A symmetric diagonally dominant matrix with
positive diagonal entries is positive definite
If a matrix is positive definite



All the diagonal entries are positive
The largest (in magnitude) element of the whole
matrix must lie on the diagonal
32
Positive Definitiness Check
20 12 25
12 15 2 


 25 2 5 



Not positive definite
Largest element is not on the diagonal

5
20 12
12 − 15 2 


5
2 25



Not positive definite
All diagonal entries are not positive

20 12 5 
12 15 2 


 5 2 25



Positive definite
Symmetric, diagonally dominant, all
diagonal entries are positive
33
Positive Definitiness Check
20 12 5 
12 15 2 


 8 2 25



A decision can not be made just by investigating
the matrix.
The matrix is diagonally dominant and all diagonal
entries are positive but it is not symmetric.
To decide, check if all the eigenvalues are positive
34
Example (Gauss-Seidel Iteration)
 4 − 1 1  x1   7 
 4 − 8 1  x  = − 21

 2  

− 2 1 5  x3   15 

  


0 
x 0 = 0 
 
0 
 

b − Ax 0

2

= 26.7395

Diagonally dominant matrix
0
0
7
7 + x2 − x3
= = 1.75
x =
4
4
1
0
21 + 4 x1 + x3 21 + 4 × 1.75
1
=
= 3 .5
x2 =
8
8
1
15 + 2 x1 − x1 15 + 2 × 1.75 − 3.5
1
2 =
= 3 .0
x3 =
5
5
1
1

b − Ax1 = 3.0414
2

b − Ax1 = 10.0452
2

Jacobi iteration
35
Example 1 continued...
1
7 + 3.5 − 3
7 + x1 − x3
2
x =
=
= 1.875
4
4
1
21 + 4 x12 + x3 21 + 4 × 1.875 + 3
2
x2 =
=
= 3.9375
8
8
2
15 + 2 x12 − x2 15 + 2 × 1.875 − 3.9375
2
=
= 2.9625
x3 =
5
5
2
1

b − Ax 2
b − Ax 2

2

2

= 0.4765
= 6.7413

Jacobi iteration

When both Jacobi and Gauss-Seidel iterations
converge, Gauss-Seidel converges faster

36
Convergence of SOR method





If 0<ω<2, SOR method converges for any initial
vector if A matrix is symmetric and positive
definite
If ω>2, SOR method diverges
If 0<ω<1, SOR method converges but the
convergence rate is slower (deceleration) than
the Gauss-Seidel method.
37
Operation count








The operation count for Gaussian Elimination
or LU Decomposition was 0 (n3), order of n3.
For iterative methods, the number of scalar
multiplications is 0 (n2) at each iteration.
If the total number of iterations required for
convergence is much less than n, then iterative
methods are more efficient than direct
methods.
Also iterative methods are well suited for
sparse matrices
38

Convergence Criteria

  • 1.
  • 2.
    Iterative Solution Methods    Startswith an initial approximation for the solution vector (x0) At each iteration updates the x vector by using the sytem Ax=b During the iterations A, matrix is not changed so sparcity is preserved  Each iteration involves a matrix-vector product  If A is sparse this product is efficiently done 2
  • 3.
    Iterative solution procedure     Writethe system Ax=b in an equivalent form x=Ex+f (like x=g(x) for fixed-point iteration) Starting with x0, generate a sequence of approximations {xk} iteratively by x k+1 =Ex k +f Representation of E and f depends on the type of the method used But for every method E and f are obtained from A and b, but in a different way 3
  • 4.
    Convergence     As k→∞, thesequence {xk} converges to the solution vector under some conditions on E matrix This imposes different conditions on A matrix for different methods For the same A matrix, one method may converge while the other may diverge Therefore for each method the relation between A and E should be found to decide on the convergence 4
  • 5.
    Different Iterative methods    JacobiIteration Gauss-Seidel Iteration Successive Over Relaxation (S.O.R)   SOR is a method used to accelerate the convergence Gauss-Seidel Iteration is a special case of SOR method 5
  • 6.
    Jacobi iteration a11 x1+ a12 x2 +  + a1n xn = b1 a21 x1 + a22 x2 +  + a2 n xn = b2  an1 x1 + an 2 x2 +  + ann xn = bn  x10   0 0  x2  x =   0  xn    1 0 0 1  (b1 − a12 x2 −  − a1n xn ) k +1 xi = bi − a11 aii  1 0 0 x1 = (b2 − a21 x10 − a23 x3 −  − a2 n xn ) 2 a22 1 0 0 x1 = (bn − an1 x10 − an 2 x2 −  − ann −1 xn −1 ) n ann 1 x1 =  ∑ aij x − j∑1aij x  j =1 =i +  i −1 k j n k j 6
  • 7.
    xk+1=Exk+f iteration forJacobi method A can be written as A=L+D+U (not decomposition) 0 0  a11 0 0  0 a12 a13   a11 a12 a13   0 a a22 a23  =  a21 0 0 +  0 a22 0  + 0 0 a23   21         a31 a32 a33   a31 a32 0  0 0 a33  0 0 0         Ax=b ⇒ (L+D+U)x=b 1 k +1 xi = aii Dxk+1 i −1 n   k k bi − ∑ aij x j − ∑ aij x j  j =1 j = i +1    Lx k  Uxk Dxk+1 =-(L+U)xk+b xk+1=-D-1(L+U)xk+D-1b E=-D-1(L+U) f=D-1b 7
  • 8.
    Gauss-Seidel (GS) iteration Usethe latest update a11 x1 + a12 x2 +  + a1n xn = b1 a21 x1 + a22 x2 +  + a2 n xn = b2  an1 x1 + an 2 x2 +  + ann xn = bn 1 0 0 1  (b1 − a12 x2 −  − a1n xn ) k +1 xi = bi − a11 aii  1 1 1 0 0 x2 = (b2 − a21 x1 − a23 x3 −  − a2 n xn ) a22 1 1 x1 = (bn − an1 x1 − an 2 x1 −  − ann −1 x1 −1 ) n 2 n ann 1 x1 =  x10   0 0  x2  x =   0  xn    i −1 ∑a x j =1 ij k +1 j  − ∑ aij x  j = i +1  n k j 8
  • 9.
    x(k+1)=Ex(k)+f iteration forGauss-Seidel Ax=b ⇒ (L+D+U)x=b k +1 i x Dx 1 = aii k+1 i −1 n   k +1 k bi − ∑ aij x j − ∑ aij x j  j =1 j = i +1    Lx k +1  (D+L)xk+1 =-Uxk+b Uxk xk+1=-(D+L)-1Uxk+(D+L)-1b E=-(D+L)-1U f=-(D+L)-1b 9
  • 10.
    Comparison    Gauss-Seidel iteration convergesmore rapidly than the Jacobi iteration since it uses the latest updates But there are some cases that Jacobi iteration does converge but Gauss-Seidel does not To accelerate the Gauss-Seidel method even further, successive over relaxation method can be used 10
  • 11.
    Successive Over RelaxationMethod  GS iteration can be also written as follows k +1 i x 1 =x + aii k i i −1 n   k +1 k bi − ∑ aij x j − ∑ aij x j  j =1 j =i   xik +1 = xik + δ ik Correction term ωδ i2 xi3 xi2 xi1 0 i x δ ωδ i1 2 i δ i1 δ i0 Multiply with ω >1 Faster convergence ωδ i0 11
  • 12.
    SOR xik +1 =xik + ωδ ik i −1 n   k +1 k x bi − ∑ aij x j − ∑ aij x j  j =1 j =i   i −1 n  1  k +1 k k +1 k xi = (1 − ω ) xi + ω bi − ∑ aij x j − ∑ aij x j  aii  j =1 j =i +1  k +1 i 1 = x +ω aii k i 1<ω<2 over relaxation (faster convergence) 0<ω<1 under relaxation (slower convergence) There is an optimum value for ω Find it by trial and error (usually around 1.6) 12
  • 13.
    x(k+1)=Ex(k)+f iteration forSOR 1 k +1 k xi = (1 − ω ) xi + ω aii i −1 n   k +1 k bi − ∑ aij x j − ∑ aij x j  j =1 j = i +1   Dxk+1=(1-ω)Dxk+ωb-ωLxk+1-ωUxk (D+ ωL)xk+1=[(1-ω)D-ωU]xk+ωb E=(D+ ωL)-1[(1-ω)D-ωU] f= ω(D+ ωL)-1b 13
  • 14.
    The Conjugate GradientMethod d 0 = r0 = b − Ax0 T ri ri αi = T d i Ad i xi +1 = xi + α i Ad i • Converges if A is a symmetric positive definite matrix • Convergence is faster T i +1 i +1 T i i r r βi +1 = r r d i +1 = ri +1 + βi +1d i 14
  • 15.
    Convergence of IterativeMethods ˆ Define the solution vector as x k Define an error vector as e ˆ x =e +x k k Substitute this into x k +1 = Ex + f k ˆ ˆ ˆ e k +1 + x = E (e k + x) + f = Ex + f + Ee k e k +1 = Ee k = EEe k −1 = EEEe k − 2 = E ( k +1) e 0 15
  • 16.
    Convergence of IterativeMethods iteration e k +1 = E ( k +1) 0 e ≤ E ( k +1) e 0 power The iterative method will converge for any initial iteration vector if the following condition is satisfied Convergence condition Lim e k +1 → 0 if k →∞ Lim E ( k +1) → 0 k →∞ 16
  • 17.
    Norm of avector A vector norm should satisfy these conditions x ≥ 0 for every nonzero vector x x = 0 iff x is a zero vector αx = α x for scalar α x+ y ≤ x + y Vector norms can be defined in different forms as long as the norm definition satisfies these conditions 17
  • 18.
    Commonly used vectornorms Sum norm or ℓ1 norm x 1 = x1 + x2 +  + xn Euclidean norm or ℓ2 norm 2 2 x 2 = x12 + x2 +  + xn Maximum norm or ℓ∞ norm x ∞ = max i xi 18
  • 19.
    Norm of amatrix A matrix norm should satisfy these conditions A ≥0 A = 0 iff A is a zero matrix for scalar α αA = α A A+ B ≤ A + B Important identitiy Ax ≤ A x x is a vector 19
  • 20.
    Commonly used matrixnorms Maximum column-sum norm or ℓ1 norm m A 1 = max ∑ aij 1≤ j ≤ n i =1 Spectral norm or ℓ2 norm A 2 = maximum eigenvalue of AT A Maximum row-sum norm or ℓ∞ norm n A ∞ = max ∑ aij 1≤i ≤ m j =1 20
  • 21.
    Example  Compute the ℓ1and ℓ∞ norms of the matrix 3 9 5 7 2 4    6 8 1    16 19 17 = A ∞ 13 15 10 = A1 21
  • 22.
    Convergence condition lim e k→∞ k +1 → 0 if lim E k →∞ ( k +1) →0 Express E in terms of modal matrix P and Λ Λ:Diagonal matrix with eigenvalues of E on the diagonal E = PΛP −1 E ( k +1) = PΛP −1 PΛP −1  PΛP −1 E ( k +1) = PΛ( k +1) P −1 k λ1 +1    λk +1 2  Λk +1 =       λk +1   n   lim E ( k +1) → 0 ⇒ lim PΛ( k +1) P −1 → 0 ⇒ lim Λ( k +1) → 0 k →∞ k →∞ k →∞ ⇒ lim λki +1 → 0 ⇒ λ i < 1 for i = 1,2 ,...,n k →∞ 22
  • 23.
    Sufficient condition forconvergence If the magnitude of all eigenvalues of iteration matrix E is less than 1 than the iteration is convergent It is easier to compute the norm of a matrix than to compute its eigenvalues Ex = λx Ex = λ x   ⇒ λ x ≤ E x ⇒ λ ≤ E Ex ≤ E x   E < 1 is a sufficient condition for convergence 23
  • 24.
    Convergence of Jacobiiteration E=-D-1(L+U)   0   − a21  a22 E=      an1 − a  nn a12 − a11 0   a23 − a22         ann −1 − ann a1n  − a11   a2 n  − a22    an −1n  −  an −1n −1   0   24
  • 25.
    Convergence of Jacobiiteration Evaluate the infinity(maximum row sum) norm of E E n ∞ <1⇒ ∑ j =1 i≠ j aij < 1 for i = 1,2,..., n aii n ⇒ aii > ∑ aij j =1 i≠ j Diagonally dominant matrix If A is a diagonally dominant matrix, then Jacobi iteration converges for any initial vector 25
  • 26.
    Stopping Criteria  Ax=b  At anyiteration k, the residual term is rk=b-Axk  Check the norm of the residual term ||b-Axk||  If it is less than a threshold value stop 26
  • 27.
    Example 1 (JacobiIteration)  4 − 1 1  x1   7   4 − 8 1  x  = − 21   2    − 2 1 5  x3   15       0  x 0 = 0    0    b − Ax 0 2 = 26.7395 Diagonally dominant matrix 0 0 7 + x2 − x3 7 x = = = 1.75 4 4 0 21 + 4 x10 + x3 21 x1 = = = 2.625 2 8 8 0 15 + 2 x10 − x2 15 1 x3 = = = 3.0 5 5 1 1 b − Ax1 = 10.0452 2 27
  • 28.
    Example 1 continued... 1 7+ x1 − x3 7 + 2.625 − 3 2 x = = = 1.65625 4 4 1 1 21 + 4 x1 + x3 21 + 4 × 1.75 + 3 2 x2 = = = 3.875 8 8 1 15 + 2 x1 − x1 15 + 2 × 1.75 − 2.625 2 2 = = 4.225 x3 = 5 5 2 1 7 + 3.875 − 4.225 = 1.6625 4 21 + 4 ×1.65625 + 4.225 3 x2 = = 3.98125 8 15 + 2 × 1.65625 − 3.875 3 x3 = = 2.8875 5 b − Ax 2 2 = 6.7413 x13 = b − Ax 2 2 = 1.9534 Matrix is diagonally dominant, Jacobi iterations are converging 28
  • 29.
    Example 2 − 21 5  x1   15   4 − 8 1  x  = − 21   2     4 − 1 1  x3   7       0  x 0 = 0    0    b − Ax 0 2 = 26.7395 The matrix is not diagonally dominant 0 0 − 15 + x2 + 5 x3 − 15 x = = = −7.5 2 2 0 21 + 4 x10 + x3 21 1 x2 = = = 2.625 8 8 1 0 x3 = 7 − 4 x10 + x2 = 7.0 1 1 b − Ax1 = 54.8546 2 29
  • 30.
    Example 2 continued... −15 + 2.625 + 5 × 7 = 11.3125 2 21 − 4 × 7.5 + 7 x1 = = −0.25 2 8 1 x3 = 7 + 4 × 7.5 + 2.625 = 39.625 1 x1 = b − Ax 2 2 = 208.3761 The residual term is increasing at each iteration, so the iterations are diverging. Note that the matrix is not diagonally dominant 30
  • 31.
    Convergence of Gauss-Seidel iteration    GSiteration converges for any initial vector if A is a diagonally dominant matrix GS iteration converges for any initial vector if A is a symmetric and positive definite matrix Matrix A is positive definite if xTAx>0 for every nonzero x vector 31
  • 32.
    Positive Definite Matrices    Amatrix is positive definite if all its eigenvalues are positive A symmetric diagonally dominant matrix with positive diagonal entries is positive definite If a matrix is positive definite   All the diagonal entries are positive The largest (in magnitude) element of the whole matrix must lie on the diagonal 32
  • 33.
    Positive Definitiness Check 2012 25 12 15 2     25 2 5    Not positive definite Largest element is not on the diagonal 5 20 12 12 − 15 2    5 2 25   Not positive definite All diagonal entries are not positive 20 12 5  12 15 2     5 2 25   Positive definite Symmetric, diagonally dominant, all diagonal entries are positive 33
  • 34.
    Positive Definitiness Check 2012 5  12 15 2     8 2 25   A decision can not be made just by investigating the matrix. The matrix is diagonally dominant and all diagonal entries are positive but it is not symmetric. To decide, check if all the eigenvalues are positive 34
  • 35.
    Example (Gauss-Seidel Iteration) 4 − 1 1  x1   7   4 − 8 1  x  = − 21   2    − 2 1 5  x3   15       0  x 0 = 0    0    b − Ax 0 2 = 26.7395 Diagonally dominant matrix 0 0 7 7 + x2 − x3 = = 1.75 x = 4 4 1 0 21 + 4 x1 + x3 21 + 4 × 1.75 1 = = 3 .5 x2 = 8 8 1 15 + 2 x1 − x1 15 + 2 × 1.75 − 3.5 1 2 = = 3 .0 x3 = 5 5 1 1 b − Ax1 = 3.0414 2 b − Ax1 = 10.0452 2 Jacobi iteration 35
  • 36.
    Example 1 continued... 1 7+ 3.5 − 3 7 + x1 − x3 2 x = = = 1.875 4 4 1 21 + 4 x12 + x3 21 + 4 × 1.875 + 3 2 x2 = = = 3.9375 8 8 2 15 + 2 x12 − x2 15 + 2 × 1.875 − 3.9375 2 = = 2.9625 x3 = 5 5 2 1 b − Ax 2 b − Ax 2 2 2 = 0.4765 = 6.7413 Jacobi iteration When both Jacobi and Gauss-Seidel iterations converge, Gauss-Seidel converges faster 36
  • 37.
    Convergence of SORmethod    If 0<ω<2, SOR method converges for any initial vector if A matrix is symmetric and positive definite If ω>2, SOR method diverges If 0<ω<1, SOR method converges but the convergence rate is slower (deceleration) than the Gauss-Seidel method. 37
  • 38.
    Operation count     The operationcount for Gaussian Elimination or LU Decomposition was 0 (n3), order of n3. For iterative methods, the number of scalar multiplications is 0 (n2) at each iteration. If the total number of iterations required for convergence is much less than n, then iterative methods are more efficient than direct methods. Also iterative methods are well suited for sparse matrices 38