Scribd
Upload
65 views6 pages

System of Linear EQ

The document describes several methods for solving systems of linear equations, including Gauss elimination, partial pivoting, LU decomposition, and Gauss-Jordan elimination. It provides pseudo code for the key steps of each algorithm, such as elimination, back substitution, pivoting, and decomposition. The algorithms are explained at a high-level, along with notes on their complexity and how they transform the original system of equations.

Uploaded by

Noushad Sojib
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
65 views6 pages

System of Linear EQ

The document describes several methods for solving systems of linear equations, including Gauss elimination, partial pivoting, LU decomposition, and Gauss-Jordan elimination. It provides pseudo code for the key steps of each algorithm, such as elimination, back substitution, pivoting, and decomposition. The algorithms are explained at a high-level, along with notes on their complexity and how they transform the original system of equations.

Uploaded by

Noushad Sojib
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

System of linear equations

Gauss Elimination:
Elimination:

Pseudo code: 1. For k=1 to n-1 // processing column (1st row without 1st column 0) 2. For i=k+1 n // for every row 3. Factor=A i,k/A k,k // find pivot 4. For j=k+1 to n // subtract 1st row from ith row after dividing by pivot 5. A i,j=A i,j factor*A k,j 6. B i=B i- factor*B k // subtract constant after dividing by pivot

Back Substitution:
Pseudo code: 1. X n=B n/A n,n 2. For i=n-1 to 1 3. Sum=B i 4. For j=i+1 to n 5. Sum=Sum-A i,j*X j 6. X i=Sum/A i,i

Nave Gauss Elimination:


1. Elimination 2. Back Substitution

Partial Pivoting:

Process: 1. Find largest element of processing column 2. Swap the two column 3. Swap value of b at this two columns Pseudo code: 1. P=k 2. Big=|A k,k| 3. For i=k+1 to n 4. Temp=|a i,k| 5. If(Temp>Big) 6. Big=Temp 7. P=I //starting column

// largest value found

8. If(p!=k) 9. For j=k to n 10. Temp=A p,j 11. A p,j=A k,j 12. A k,j=Temp 13. Temp=Bp 14. Bp=Bk 15. Bk=Temp

// is largest value is the first one? // if not, then swap the coefficient of X, A[]

//swap finished //swap the constants, B[] //Finished pivoting

Scaling:
1. The maximum coefficient in each row can be 1 2. To do: divide all of the coefficient in a row with its largest coefficient

Pseudo code: 1. 2. 3. 4. 5. 6. 7. 8. 9. K Big=A k,k For(i=k+1 to n) Temp=A i,k If(Temp>Big) Big=Temp For(i=k to n) A k,i=A k,i/Big B k=B k/Big

** to perform better scaling should be embedded with elimination process

Gauss Elimination with Partial Pivoting:

** The equations are not scaled, but scaled values of the elements are used to determine whether pivoting is to be implemented.

Gauss(A, B, n, X, tol, er) 1. Arry S[n] 2. er=0 3. For i=1 to n // find the largest value in each column 4. S i=abs(A i,1) 5. For j=2 to n 6. If(abs(A i,j)>S i) 7. S i=abs(A i,j) 8. Call Eliminate(A, S, n, B, tol, er) // now eliminate to make upper triangle 9. If(er!=-1) // terminate condition (detect near zero) 10. Call Substitute(A, n, B, X)

Eliminate(A, S, n, B, X) 1. For k=1 to n-1 // processing column (ith row without 1st column 0)

2. Call Pivot(A, B, S, n, k) // find pivot 3. if( abs(A k,k/S k)< tol // in scaling if round-off error exceeds tolerance then return -1 and exit 4. Er=-1 5. Exit 6. For i=k+1 to n //for every row 7. factor=A i,k/A k,k 8. For j=k+1 to n // subtract ith row from the processing row after dividing by pivot. 9. A i,j=A i,j factor*A k,j 10. B i=B i factor*B k // subtract the constant after dividing by pivot 11. if(abs(A k,k/S k)< tol) // in scaling if error exceeds tolerance then return -1 and exit 12. er= -1

Pivot(A, B, S, n, k) 1. P=k 2. Big=abs(A k,k/S k) 3. For i=k+1 to n 4. Temp=abs(A i,k/S i) 5. If(Temp>Big) 6. Big=Temp 16. P=i 7. 8. If p!=k 9. For j=k to n 10. Temp=A p,j 11. A p,j=A k,j 12. A k,j=Temp 13. Temp=B p 14. B p=B k 15. B k=Temp 16. Temp= S p 17. S p=S k 18. S k=Temp

// assume first is largest with scaling // go with scaling

// largest value found with scaling

Substitute (A, n, B, X) 1. X n=B n/A n,n 2. For i=n-1 to 1 3. Sum=0 4. For j=i+1 to n 5. Sum=Sum+A i,j*X j 6. X i=(B i-Sum) / A i,i

Gauss Jordan:
Gauss-Jordan is a variation of Gauss Elimination. Here equation is eliminated from all other equations rather than the subsequent ones like Gauss elimination. Then there makes a identity matrix. So no back substitution is required to obtain solution. The complexity is same as Gauss elimination.

Process:

1. 2. 3. 4.

Partial pivoting before normalize Normalize first row (make 1) Find pivot Eliminate

Pseudo code: 1. For k=1 to n-1 2. Normalize() // make first column=1 (to make identity matrix) 3. For i = 1 to n 4. If( i!=k) 5. Factor=A i,k 6. For j=1 to n 7. A i,j=A i,j - factor*A k,j 8. B i=B i-factor*B k

LU Decomposition:
[A]{X} = {B} [A]{X} {B} = 0 After Gauss elimination [U]{X} {D}=0 1 Suppose [L] = 0 0 That has property that, [L]{ [U]{X} {D} } = [A]{X} {B} So , [L][U] = [A] [L]{D} = {B} 0 1 0 0 0 1

Where, 1 [L] = 0 0 0 ,n=3

f21 1

f31 f32 1

Pseudo code:

Decompose(A, n) 1. For k=1 to n-1 2. For i=k+1 to n 3. Factor=A i,k/A k,k 4. A i,k=Factor 5. For j=k+1 to n 6. A i,j=A i,j Factor*A k,j

Substitute(A, n, B, X) 1. For i=2 to n // forward substitution 2. Sum=B i 3. For j=1 to i-1 4. Sum=sum-A i,j * B j 5. B i=Sum // {D} found 6. 7. X n=B n/A n,n // back substitute (find {X}) 8. For i=n-1 to -1 9. Sum=0 10. For j=i+1 to n 11. Sum=sum+A I,j *X j 12. X i=(B i sum)/A I,i

LU-Decomposition Pseudo code:


** similar with Gauss Elimination with partial pivoting algorithm, except swapping rows but ordering them. LUDecomp(A, B, n, tol, X, Er) 1. 2. 3. 4. 5. Arry O[n],S[n] Er=0 Call Decompose(A, n, tol, O, S, Er) if Er!= 1 call substitute(A, O, n, B, X)

Decompose(A, n, tol, O, S, Er) 1. For i=1 to n 2. O i=i 3. S i=abs(A i,1) 4. For j=2 to n 5. If (abs (Ai,j)>S i) 6. S i= abs(A i,j) 7. For k=1 to n-1 8. Pivot(A, O, S, n, k) 9. If( abs(A o(k),k/S o(k) < tol 10. Er=-1 11. Print A o(k),k/S o(k) 12. Exit 13. For i=k+1 to n 14. Factor = A o(i),k/A o(k),k 15. A o(i),k = Factor

16. For j=k+1 to n 17. A o(i),j = A o(i),j Factor * A o(k),j 18. If(abs (A o(k),k/S o(k) < tol 19. Er= - 1 20. Print A o(k),k/S o(k) Pivot(A, O, S, n, k) 1. P=k 2. Big=abs(A o(k),k/ S o(k)) 3. For i=k+1 to n 4. Temp=abs(A o(i),k/ S o(i)) 5. if( Temp> Big) 6. Big= Temp 7. P=i 8. Temp=O p 9. O p=O k 10. O k=Temp Substitute(A, O, n, B, X) 1. For i=2 to n 2. Sum=B o(i) 3. For j=1 to i-1 4. Sum=sum- A o(i),j *B o(j) 5. B o(i)=sum 6. X n=B o(n)/A o(n),n 7. For i=n-1 to -1 8. Sum=0 9. For j=i+1 to n 10. Sum=sum+ A o(i),j * X j 11. X i=( B o(i) Sum)/ A o(i),j

You might also like