Scribd
Upload
32 views

Typed Lecture For Sec 7 - 1

This document defines and provides examples of vector and matrix norms, which are used to measure distances between vectors and matrices. It introduces common vector norms (Euclidean, 1-norm, infinity norm) and defines a matrix norm. A key theorem states that for any vector norm and its induced matrix norm, the norm of a matrix-vector product is always less than or equal to the matrix norm times the vector norm. Examples are provided to illustrate these concepts.

Uploaded by

jtownball
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
32 views

Typed Lecture For Sec 7 - 1

This document defines and provides examples of vector and matrix norms, which are used to measure distances between vectors and matrices. It introduces common vector norms (Euclidean, 1-norm, infinity norm) and defines a matrix norm. A key theorem states that for any vector norm and its induced matrix norm, the norm of a matrix-vector product is always less than or equal to the matrix norm times the vector norm. Examples are provided to illustrate these concepts.

Uploaded by

jtownball
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/ 5

Section 7.

1: Norms (vector norms and matrix norms)


The two main themes in chapter 7 are: 1) iterative methods for solving linear systems of equations; and 2) analysis of the error in a computed solution to a linear system of equations. For both of these topics, we need to have some way to measure the distance between two vectors. For the error analysis, the error is given in terms of a distance between the computed solution and the true solution. For numerical solutions found using iterative methods, we will want to find the distance between consecutive approximate solutions. We use norms to define distances. Notations:

denotes the set of all real numbers n denotes the set of all n-dimensional real vectors (we will usually use column vectors) f : D R means f is a function with domain D and range contained in the set R w.r.t. stands for with respect to

r r Definition: A function : n is a vector norm if for all x , y n and all , r 1) x 0 r r r 2) x = 0 if and only if x = 0 r r 3) x = x r r r r 4) x+y x + y (this property is called the triangle inequality).
In the above definition, is used as a place-holder for a single vector from n . r r We often think of x as the length of x w.r.t. . Examples of vector norms:

x1 r x2 Let x = n . The most commonly used norms on n are: M xn


1) Euclidean norm (or 2-norm):

r x

n = xi2 i =1

2)

1-norm:

r x

= xi
n i =1

3)

-norm:

r x

max x i = 1,2, L , n i

Section 7.1

Page 1

r r r T For example, if x = [3 5 2 1] , then x 1 = 11 , x 2 = 39 6.245 , and r x = 5.

Try this problem: 7.1.1


a) b) c

r T For x = [ 3 7 2 1 1 ] , find r x 1 r x 2 r x .

The above vector norms are special cases of a more general type of vector r T norm called a p-norm: if p 1 , then the p-norm of x = [x1 x 2 L x n ] is
r x
p

n p = xi i =1

r (it turns out that x

r lim x p

).

r r Definition: The distance between vectors x & y w.r.t. vector norm is r r r r d (x, y) = x y .
3 3 r r The distance between x = 2 and y = 1 is: 4 5 r r d ( x , y ) = 2 w.r.t. 1 r r d ( x , y ) = 2 w.r.t. 2 r r d ( x , y ) = 1 w.r.t.

Example: 1) 2)
3)

Note that the distance depends on the norm being used.

r r Definition: A sequence {x ( k ) }k =1 in n converges to x n w.r.t. if

r r lim x ( k ) x = 0 . k

Theorem: In n , convergence w.r.t. any norm implies convergence w.r.t. every norm. (We usually just talk about convergence in n instead of convergence in n w.r.t. a r r r r particular norm. If {x ( k ) }k =1 converges to x n , we write lim x ( k ) = x .) k

Section 7.1

Page 2

Theorem: In n , convergence w.r.t.


r converges to x n w.r.t.

r is component-wise convergence, i.e. x ( k )

{ }

k =1

if and only if

r ( xi is the i th component of x and xi(k )


Example: Using the infinity norm,

lim x ( k ) = x for each i = 1, 2, L , n . i k i r is the i th component of x (k ) ).

k k lim k k k +1 + 1 1 lim cos k = lim cos k = 0 k k k k 1 k k 5 e e lim k 5e + 2k k 5e k + 2k Definition: A matrix norm for the set of nn matrices is a real valued function defined for all nn matrices such that for all nn matrices A and B, and for all :
1) 2) 3) 4) 5)

A 0 A = 0 if and only if A = 0 nn A = A A+ B A + B AB A B

Note that we require one more condition for a matrix norm than we did for a vector norm.

Definition: If is a vector norm on n , then the corresponding natural matrix norm,


r max r or induced matrix norm, for nn matrices is given by A = x = 1 Ax for each
nn matrix A.

Comments:

1)

It would actually be a theorem that the above process produces something that satisfied the conditions for a matrix norm. Although we use the same notation for a vector norm and its induced matrix norm, we can generally tell from context whether we are working with a vector norm or a matrix norm. The above definition is not very convenient for actually calculating matrix norms.

2)

3)

Section 7.1

Page 3

Theorem: If A is an nn matrix, then


1)
A

max a 1 j n ij i =1
n

2)

max a 1 i n ij j =1
n

(There is a formula for A 2 , but it is not very convenient. We will see the formula in section 7.2).

3 5 2 1 5 9 Example: If A = 3 0 1 1 1 2

6 2 , then 3 7

= max {8, 11, 14, 18} = 18 and

= max { , 17, 7, 11} = 17 . 16

Try this problem: 7.1.2

a) b)

2 3 5 For A = 7 0 1 , find 6 1 1 A 1 A

. and AT

Note that in general, we will have AT

= A

= A 1 . Thus, in

some examples, the infinity norm of a matrix is larger than the one norm, and in other examples, the one norm is larger than the infinity norm. The following theorem will be used when we analyze the errors of computed solutions to linear systems. That error analysis is in section 7.5 of the text.

Theorem: If represents both a vector norm on n and the corresponding induced r matrix norm for nn matrices, then for every x n and every nn matrix A, r r Ax A x .

Section 7.1

Page 4

Example:

3 2 5 0 1 1 If A = 2 4 5 1 7 3

1 2 3 r r and x = 1 , then Ax = 1 6 1 3
1

6 7 . We have 23 3

r Ax

= 39 14 7 = A

r x

and

r Ax

= 23 17 3 = A

r x

Try this problem: 7.1.3

5 2 1 2 r r r 3 and x = 3 , verify that both Ax A x and For A = 3 1 1 1 1 6 4 3 6 r r r r Ax A x . Given that A 2 9.2682 , verify that Ax 2 A 2 x 2 .

Section 7.1

Page 5

You might also like