Lecture2 Hand

Lecture 2 The rank of a matrix
Eivind Eriksen
BI Norwegian School of Management Department of Economics
September 3, 2010
Eivind Eriksen (BI Dept of Economics)
September 3, 2010
1 / 24
Linear dependence
Linear dependence
To decide if a set of m-vectors {a1 , a2 , . . . , an } are linearly independent, we have to solve the vector equation x1 a1 + x2 a2 + + xn an = 0 We know that x = 0 is one solution, the trivial solution. Are there other (non-trivial) solutions? If yes, then the vectors are linearly dependent. We can use a non-trivial solution to express one vector as a linear combination of the others. If no, then the vectors are linearly independent
September 3, 2010
2 / 24
Linear dependence
Linear systems and vector equations
A linear system of m equations is the same as a single vector equation of m-vectors. We may therefore re-write a vector equation as a linear system, and also re-write a linear system as a vector equation. Example Write the following linear system as a vector equation: 2x1 4x1 + 2x2 6x2 + x3 2x3 3x3 = = = 0 0 0
September 3, 2010
3 / 24
Linear dependence
Linear systems and vector equations
Solution We re-write the three equations as one equation of 3-vectors: 2 2 1 0 x1 4 + x2 0 + x3 2 = 0 0 6 3 0 We may write this as x1 a1 + x2 a2 + x3 a3 = 0 Note that a1 , a2 , a3 are the columns of the coecient matrix of the linear system, and 0 is the last (augmented) column of the augmented matrix.
September 3, 2010
4 / 24
Linear dependence
Criterion for linear independence
Theorem Let a1 , a2 , . . . , an be n-vectors, and let A be the n n matrix with these vectors as columns. Then {a1 , a2 , . . . , an } are linearly independent if and only if a11 a12 . . . a1n a21 a22 . . . a2n det(A) = . . . =0 .. . . . . . . . an1 an2 . . . ann Idea for proof: The linear system Ax = 0 has a unique solution (that is, only the trivial solution) if and only if det(A) = 0.
September 3, 2010
5 / 24
Linear dependence
An example
Example Show that a1 , a2 , a3 are linearly independent when 2 2 1 a1 = 4 , a2 = 0 , a3 = 2 0 6 3 Solution Since we have 2 2 1 4 0 2 = (4) 0 + (2) 12 = 24 = 0 0 6 3 it follows that the vectors are linearly independent.
Eivind Eriksen (BI Dept of Economics) Lecture 2 The rank of a matrix September 3, 2010 6 / 24
The rank of a matrix
Rank of a matrix
Let A be any m n matrix. Then A consists of n column vectors a1 , a2 , . . . , an , which are m-vectors. Denition The rank of A is the maximal number of linearly independent column vectors in A, i.e. the maximal number of linearly independent vectors among {a1 , a2 , . . . , an }. If A = 0, then the rank of A is 0. We write rk(A) for the rank of A. Note that we may compute the rank of any matrix square or not.
September 3, 2010
7 / 24
Rank of 2 2 matrices
Let us rst see how to compute the rank of a 2 2 matrix: Example The rank of a 2 2 matrix A =
a b c d
is given by:
rk(A) = 2 if det(A) = ad bc = 0, since both column vectors are independent in this case
0 rk(A) = 1 if det(A) = 0 but A = 0 = ( 0 0 0 ), since both column vectors are not linearly independent, but there is a single column vector that is linearly independent (i.e. non-zero)
rk(A) = 0 if A = 0 How do we compute rk(A) for an m n matrix A?
September 3, 2010
8 / 24
Computing rank using Gauss elimination

Gauss elimination Use elementary row operations to reduce A to echelon form. The rank of A is the number of pivots or leading coecients in the echelon form. In fact, the pivot columns (i.e. the columns with pivots in them) are linearly independent. Note that it is not necessary to nd the reduced echelon form any echelon form will do since only the pivots matter. Possible ranks Counting possible number of pivots, we see that rk(A) m and rk(A) n for any m n matrix A.
September 3, 2010
9 / 24
Rank: Example using Gauss elimination

Example Find the rank of the matrix 1 0 2 1 A = 0 2 4 2 0 2 2 1 Solution We use elementary row 1 A = 0 0
operations: 0 2 1 2 4 2 2 2 1
1 0 2 1 0 2 4 2 0 0 2 1
Since the echelon form has pivots in the rst three columns, A has rank rk(A) = 3. The rst three columns of A are linearly independent.
Computing rank using determinants

Denition Let A be an m n matrix. A minor of A of order k is a determinant of a k k sub-matrix of A. We obtain the minors of order k from A by rst deleting m k rows and n k columns, and then computing the determinant. There are usually many minors of A of a given order. Example Find the minors of order 3 of the matrix 1 0 2 1 A = 0 2 4 2 0 2 2 1
September 3, 2010
11 / 24
Computing minors
Solution We obtain the determinants of order 3 by keeping all the rows and deleting one column from A. So there are four dierent minors of order 3. We compute one of them to illustrate: 1 0 2 0 2 4 = 1 (4) + 2 0 = 4 0 2 2 The minors of order 3 are called the maximal minors of A, since there are no 4 4 sub-matrices of A. There are 3 6 = 18 minors of order 2 and 3 4 = 12 minors of order 1
September 3, 2010
12 / 24
Computing rank using minors

Proposition Let A be an m n matrix. The rank of A is the maximal order of a non-zero minor of A. Idea of proof: If a minor of order k is non-zero, then the corresponding columns of A are linearly independent. Computing the rank Start with the minors of maximal order k . If there is one that is non-zero, then rk(A) = k . If all maximal minors are zero, then rk(A) < k , and we continue with the minors of order k 1 and so on, until we nd a minor that is non-zero. If all minors of order 1 (i.e. all entries in A) are zero, then rk(A) = 0.
September 3, 2010
13 / 24
Rank: Examples using minors
Example Find the rank of the matrix 1 0 2 1 A = 0 2 4 2 0 2 2 1 Solution The maximal minors have order 3, and we found that the one obtained by deleting the last column is 4 = 0. Hence rk(A) = 3.
September 3, 2010
14 / 24
Example Find the rank of the matrix 1 2 1 1 A = 9 5 2 2 7 1 0 4
September 3, 2010
15 / 24

Solution The maximal minors have order 3, so we compute the 4 minors of order 3. The rst one is 1 2 1 9 5 2 = 7 (1) + (1) (7) = 0 7 1 0 The other three are also zero. Since all minors of order 3 are zero, the rank must be rk(A) < 3. We continue to look at the minors of order two. The rst one is 1 2 = 5 18 = 13 = 0 9 5 It is not necessary to compute any more minors, and we conclude that rk(A) = 2. In fact, the rst two columns of A are linearly independent.
Application: Linear independence

Example Show that the vectors are linearly independent: 1 1 a1 = 1 , a2 = 0 2 3 Solution The vectors are linearly independent if and only if rk(A) = 2, where A is the matrix with a1 and a2 as columns. Since we have 1 1 =1=0 1 0 it follows that rk(A) = 2.
Rank and linear systems
Theorem Let Ab = (A|b) be the augmented matrix of a linear system Ax = b in n unknowns. Then we have:
1 2
The linear system is consistent if and only if rk Ab = rk A. If the linear system is consistent, then it has a unique solution if and only if rk(A) = n. Moreover, if rk(A) < n, then the system has n rk(A) free variables.
Idea of proof: Think of the pivots in the reduced echelon form of the system.
September 3, 2010
18 / 24
Linear system: Example using rank
Example Is the following linear system consistent? Does it have a unique solution? 2x1 4x1 + 2x2 6x2 + x3 2x3 3x3 = = = 1 2 4
September 3, 2010
19 / 24
Linear system: Example using rank
Solution We form the matrices 2 2 1 A = 4 0 2 , 0 6 1
2 2 1 1 Ab = 4 0 2 2 0 6 1 4
We compute that det(A) = 24 = 0, so rk(A) = 3 and rk Ab = 3 (since the determinant is a maximal minor of the augmented matrix). Hence the system is consistent. In fact, n rk A = 3 3 = 0, so the system has a unique solution.
September 3, 2010
20 / 24
Linear system: Explicit solutions using minors
Interpretation of minors We consider a consistent linear system Ax = b and let k = rk(A). Then there is a non-zero minor of order k . We can interpret this minor in the following way: The deleted rows are not essential, and we may disregard them. Hence we only regard the rows (equations) that are in the minor. The variables corresponding to deleted columns represent free (independent) variables. The variables corresponding to columns in the minor are basic (dependent) variable. We may write down the solution of the system by solving the equations in the minor for the basic (dependent) variables.
September 3, 2010
21 / 24
Linear system: Example solved using minors
Example Solve the following (consistent) linear system using minors: x1 x1 x2 x2 + + + 2x3 x3 3x3 + + 3x4 3x4 = = = 0 0 0
We remark that this system is consistent, since it has the trivial solution x = 0.
September 3, 2010
22 / 24

Solution We compute the rank of the coecient matrix 1 1 2 3 A = 0 1 1 0 1 0 3 3 After some computations, we see that all maximal (order 3) minors are zero. However, the minor of order 2 obtained by deleting the last row and the last two columns from A is 1 1 =1=0 0 1 This means that rk A = 2, and that the linear system has 4 2 = 2 free variables.

Solution The free variables are x3 , x4 , and we may express x1 , x2 in terms of the free variables using the rst two equations: x1 x2 = 2x3 3x4 x2 = x3 This gives x1 = 3x3 3x4 x2 = x3 x3 = free variable x4 = free variable

Lecture2 Hand

Uploaded by

Copyright:

Available Formats

Lecture2 Hand

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture2 Hand

Uploaded by

Copyright:

Available Formats

Lecture 2 The rank of a matrix

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

Linear systems and vector equations

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

Linear systems and vector equations

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

Criterion for linear independence

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

rk(A) = 0 if A = 0 How do we compute rk(A) for an m n matrix A?

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Computing rank using Gauss elimination

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Rank: Example using Gauss elimination

The rank of a matrix

Computing rank using determinants

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Computing rank using minors

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Rank: Examples using minors

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Rank: Examples using minors

Example Find the rank of the matrix 1 2 1 1 A = 9 5 2 2 7 1 0 4

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Rank: Examples using minors

The rank of a matrix

Application: Linear independence

The rank of a matrix

Rank and linear systems

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix

The rank of a matrix

Linear system: Example using rank

Eivind Eriksen (BI Dept of Economics)

Lecture 2 The rank of a matrix