Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
0% found this document useful (0 votes)
32 views

Lin Alg Week 10 A

This document describes the Gram-Schmidt process for turning an initial basis into an orthonormal basis. It works as follows: 1) The first basis vector v1 is set to the first initial basis vector u1 normalized to have length 1. 2) Subsequent basis vectors vk are found by taking the initial basis vector uk and subtracting its projections onto previous basis vectors v1, ..., vk-1, then normalizing to have length 1. 3) After n steps, the resulting vectors v1, ..., vn form an orthonormal basis, proving that every finite-dimensional real inner product space has an orthonormal basis.

Uploaded by

Anthony Jagdeo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
32 views

Lin Alg Week 10 A

This document describes the Gram-Schmidt process for turning an initial basis into an orthonormal basis. It works as follows: 1) The first basis vector v1 is set to the first initial basis vector u1 normalized to have length 1. 2) Subsequent basis vectors vk are found by taking the initial basis vector uk and subtracting its projections onto previous basis vectors v1, ..., vk-1, then normalizing to have length 1. 3) After n steps, the resulting vectors v1, ..., vn form an orthonormal basis, proving that every finite-dimensional real inner product space has an orthonormal basis.

Uploaded by

Anthony Jagdeo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Proof: Suppose that

1 v 1 + + n v n = 0.
Orthogonal or orthonormal bases are very useful but how do we
find them? We will show that every non-zero finite-dimensional real We must show that 1 = = n = 0. Fixing i, and taking the inner
inner product space has an orthonormal basis, by giving an algorithm product of both sides of this equation with v i , we obtain
for turning an initial non-orthonormal basis into an orthonormal one.
h0, v i i = h1 v 1 + + n v n , v i i
For this we will need
= 1 hv 1 , v i i + + n hv n , v i i
Proposition 4.12: If S = {v 1 , . . . , v n } is an orthogonal set of non-zero
= i hv i , v i i
elements in a real inner product space V , then the elements of S are
linearly independent. by the orthogonality of the v i s. But

h0, v i i = h0.0, v i i = 0h0, v i i = 0

and hv i , v i i =
6 0 and so i = 0 as required.

Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 1/8 Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 2/8

Step 1: Let
u1
Theorem 4.13: Every non-zero finite-dimensional real inner product v1 = .
ku 1 k
space has an orthonormal basis.
Then kv 1 k = 1 and this will be the first vector in our new basis.
For most of the proofs thus far in the module, we have not needed the
technical details of the proof in order to be able to apply the result. Step 2: Let
w 2 = u 2 hu 2 , v 1 iv 1 .
However this Theorem is different, as our proof will consist of a
description of the method used to construct an orthonormal basis. This Claim that w 2 is orthogonal to v 1 and is non-zero.
procedure is called the Gram-Schmidt process and will be very First we prove orthogonality: We have
important in what follows.
Proof: Let V be such a space, of dimension n say, and suppose that hw 2 , v 1 i = hu 2 hu 2 , v 1 iv 1 , v 1 i
u 1 , . . . u n is a basis of V . We will construct a new, orthonormal basis = hu 2 , v 1 i hu 2 , v 1 ihv 1 , v 1 i
for V from this. = hu 2 , v 1 i hu 2 , v 1 i = 0

as v 1 has norm 1.

Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 3/8 Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 4/8

We now proceed by induction. So suppose that we have already


Now suppose that w 2 = 0. Then constructed an orthonormal set v 1 , . . . , v k 1 ; we will show how to
construct v k .
u 2 hu 2 , v 1 iv 1 = 0. Step k: Let
1
kX
But this implies that w k = uk hu k , v i iv i .
u1
u 2 = hu 2 , v 1 i i=1
ku 1 k
Claim that w k is orthogonal to v 1 , . . . , v k 1 and is non-zero.
and hence that u 1 and u 2 are not independent (which is impossible).
First we prove orthogonality: For each 1 j < k we have
Now let
w2
v2 = 1
kX
kw 2 k hw k , v j i = hu k , v j i hu k , v i ihv i , v j i
which has norm 1, and so v 1 , v 2 form an orthonormal set. i=1
= hu k , v j i hu k , v j i = 0.

Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 5/8 Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 6/8

If w k = 0 then
1
kX
After n steps we get an orthonormal set v 1 , . . . , v n . By Proposition
uk hu k , v i iv i = 0.
4.12 these are linearly independent, and so they form a basis of V by
i=1
Corollary 1.32.
Each v i is a linear combination of u 1 , . . . , u i and so we will get
Example 4.14: Take R3 with the usual scalar product, and apply
1 X
kX i Gram-Schmidt to the basis
uk aij u i = 0
i=1 j=1 u 1 = (1, 1, 1) u 2 = (0, 1, 1) u 3 = (0, 0, 1).

for some aij , which contradicts the linear independence of the u i s.


Now let Example completed by hand in the lecture.
wk
vk =
kw k k
which has norm 1, and so v 1 , . . . , v k form an orthonormal set.

Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 7/8 Anton Cox (City University London) AS2051/MA2602 Lecture 26 Autumn 2015 8/8

You might also like