Exercise09 Sol

2
- 1 - AMS 147 Computational Methods and Applications Exercise #9 1. Write out the 1-norm of vector x = x 1 , x 2 , , x N ( = T . Answer: 1-norm: x 1 = x j j =1 N 2. Write out the 2-norm of vector x = x 1 , x 2 , , x N ( = T . Answer: 2-norm: x 2 = x j 2 j =1 N 1 2 3. Write out the infinity-norm of vector x = x 1 , x 2 , , x N ( = T . Answer: -norm: x = max j x j 4. Write out the norm of matrix A in terms of vector norms. Answer: A = max x 0 Ax x 5. Write out the condition number of matrix A in terms of matrix norms. Answer: cond A ( = = A 1 A

description

Solution to exercise 9

Transcript of Exercise09 Sol

  • - 1 -

    AMS 147 Computational Methods and Applications

    Exercise #9

    1. Write out the 1-norm of vector x = x1, x2 ,, xN( )

    T.

    Answer:

    1-norm: x1

    = x jj =1

    N

    2. Write out the 2-norm of vector x = x1, x2 ,, xN( )

    T.

    Answer:

    2-norm: x2

    = x j2

    j =1

    N1

    2

    3. Write out the infinity-norm of vector x = x1, x2 ,, xN( )

    T.

    Answer:

    -norm: x = maxj

    x j

    4. Write out the norm of matrix A in terms of vector norms.

    Answer:

    A = maxx 0

    Ax

    x

    5. Write out the condition number of matrix A in terms of matrix norms.

    Answer:

    cond A( ) = A 1 A

  • AMS 147 Computational Methods and Applications

    - 2 -

    6. Consider a linear system A x = b with condition number cond(A) < 10.

    We use Gauss elimination without pivoting to solve A x = b.

    Is the relative error in the numerical solution guaranteed to be small? Why?

    Answer:

    No. The relative error in the numerical solution is not guaranteed to be small because Gauss

    elimination without pivoting is not a good numerical method.

    7. Consider a linear system A x = b with condition number cond(A) = 1020.

    We use Gauss elimination with pivoting to solve A x = b.

    Is the relative error in the numerical solution guaranteed to be small? Why?

    Answer:

    No. The relative error in the numerical solution is not guaranteed to be small because the

    condition number is very large.

    8. Let A be an N N matrix. For N = 106, what is the amount of RAM memory required for

    storing all elements of matrix A as double precision real numbers?

    Answer:

    To store all 1012 elements of matrix A as double precision real numbers, it requires

    1012 64 bits = 1012 8 Bytes = 8000 GB