You are on page 1of 17

Problem

Prove that an m × n matrix U has orthonormal columns if and only if


UT U = I .
Proof.
To simplify the notation, suppose U has only 3 columns each a vector in
R m . LetU =
 [u1 u2 u3 ]. Then 
u1T u1T u1 u1T u2 u1T u3
   
U T U = u2T  [u1 u2 u3 ] = u2T u1 u2T u2 u2T u3  Note that the columns of U
   
   
u3T u3T u1 u3T u2 u3T u3
are orthogonal if and only if u1T u2 = u2T u1 = 0;u1T u3 = u3T u1 = 0;
u2T u3 = u3T u2 = 0. Also the columns of U have unit length if and only if
u1T u1 = u2T u2 = u3T u3 = 1. Thus U T U = I . This proves the result in general.

Dr. Sabitha D’Souza October 06,2022 1 / 13


2.Let U be an n × n matrix with orthonormal columns. Show that
det(U) = ±1.
Proof.
Since U is an n × n matrix with orthonormal columns by above result we
have U T U = I . Taking determinant on both sides we get
det(U T U) = det(U T ) det(U) = det(U) det(U) = det(U)2
det(I ) = det(U)2
1 = det(U)2
This implies det(U) = ±1

Dr. Sabitha D’Souza October 06,2022 2 / 13


Dr. Sabitha D’Souza October 06,2022 3 / 13
Procedure

1 Rename
 the columns
  of matrix 
A as
 x1 , x2 , x3 .
1 0 0
     
1 1 0
     
 , x2 =   and x3 =  .
x1 =      
1 1 1
     
1 1 1
2 Use Gram Schmidth orthogonalization process to obtain orthonormal
basis {v1 , v2 , v3 }

Dr. Sabitha D’Souza October 06,2022 4 / 13


 
1
 
1
 
2 v1 = 
 

1
 
1

Dr. Sabitha D’Souza October 06,2022 5 / 13


   
−3
1
   4 
 1 
1
 
 4 
2 v1 = 
  , v2 =  1 
  
1  4 
   
1
1 4

Dr. Sabitha D’Souza October 06,2022 5 / 13


   
−3
1
   4 
 1 
1
 
 4 
2 v1 = 
  , v2 =  1 
  
1  4 
   
1
1 4
Simplify 
v2 for easier calculation. Multiply by 4.
−3
 
1
 
So v2′ = 
 

1
 
1
x3 .v1 x3 .v2 ′
Compute v3 = x3 − v1 .v1 .v1 − .v
v2′ .v2′ 2

Dr. Sabitha D’Souza October 06,2022 5 / 13


 
0
 
 −2 
 3 
2 v3 = 
 1 

 3 
 
1
3
 
0
 
−2
 

Multiply by 3. v3 =  
 
1
 
1
′ ′
Thus {v1 , v2 , v3 } is the orthogonl basis of A.

Dr. Sabitha D’Souza October 06,2022 6 / 13


3 Normalise {v1 , v2′ , v3′ } to obtain orthonormal basis {u1 , u2 , u3 }
Thus Q = [u1 u2 u3 ]
 
1 −3
√ 0
2 12 
1 −2 
 2 √1 √ 
Q= 12 6
1
 2 √1 √1 

 12 6
1 √1 √1
2 12 6

4 To find R, we use Q T Q = I
We have A = QR. Multiply by Q T on the left
Q T A = Q T QR
Q T A = IR = R. Thus R = Q T A.

Dr. Sabitha D’Souza October 06,2022 7 / 13


 
  1 0 0
1 1 1 1  
 2 2 2 2 
 1 1 0

R =  √−3 √1 √1 √1   =

4
 12 12 12 12  
1 1 1

−2 √1 1
0 √
6 6

6
 
1 1 1

Dr. Sabitha D’Souza October 06,2022 8 / 13


 
  1 0 0  
1 1 1 1 3
2 2 2 2 
  2 2 1
 1 1 0
   
R =  √−3 √1 √1 √1   = √3 √2 
  
4 0
 12 12 12 12  
1 1 1
  12 12 
−2 √1 1 √2
0 √
6 6

6
  0 0 12
1 1 1
Thus A = QR.

Dr. Sabitha D’Souza October 06,2022 8 / 13


Problem 2

 
3 2
Find QR factorization of A =  
4 5
   
3 −4 26
5
Ans: Q =  5 5 , R =  5 
4 3
5 5 0 75

Dr. Sabitha D’Souza October 06,2022 9 / 13


Problem 3

 
2 3 1
 
Find QR factorization of A = 1 1 1
 
 
2 1 2
 
7 1 1
 3 √3 3 
Ans: R =  0 √3 
 
2 2 

0 0 √1
18

Dr. Sabitha D’Souza October 06,2022 10 / 13


An application of QR factorization to solve system of equations
Suppose we are given the consistent system of linear equations Ax = b,
where A is an m × n matrix with linearly independent columns.
Let A = QR be a QR factorization of A.
Using the result that Q T Q = Im , we obtain the equivalent systems

Dr. Sabitha D’Souza October 06,2022 11 / 13


An application of QR factorization to solve system of equations
Suppose we are given the consistent system of linear equations Ax = b,
where A is an m × n matrix with linearly independent columns.
Let A = QR be a QR factorization of A.
Using the result that Q T Q = Im , we obtain the equivalent systems Ax = b
Put A = QR we get QRx = b. Multiply by Q T on both the sides.
Q T QRx = Q T b which reduces to Im Rx = Q T b
i.e Rx = Q T b.
Note that R is an upper triangular matrix and hence the system is easier
to solve.

Dr. Sabitha D’Souza October 06,2022 11 / 13


Problem

Solve the system by QR method

 √ 
1/2 1/ 2 1/2  
  2 2 5/2
1/2 0 −1/2
   √ √ 
Q= and R = −1/
 
 √  0 2 2
1/2 −1/ 2 1/2 
  
  0 0 1/2
1/2 0 −1/2

Dr. Sabitha D’Souza October 06,2022 12 / 13


 
4
 
Solve Rx = QT b . Ans: x = −2
 
 
1

Problem 2
Solve the system of equations by QR method

Ans: [x1 = 2 x2 = −1 x3 = 3]

Dr. Sabitha D’Souza October 06,2022 13 / 13

You might also like