27th December 2014 M03_LEON2218_09_SE_C03 page 182

182 Chapter 3 Vector Spaces

3. (Column Space and Reduced Row Echelon Form) (b) Consider now a new system Cx = b, where C is
Set constructed as follows:
B = round(10 ∗ rand(6, 3)) u = round(10 ∗ rand(7, 1))
v = round(10 ∗ rand(7, 1))
X = round(10 ∗ rand(3, 4))
E = u ∗ v
C =B∗X C =A+E
A = [B C] The matrices C and A differ by the rank-1 mat-
(a) How are the column spaces of B and C related? rix E. Use MATLAB to verify that the rank of E
(See Exercise 28 in Section 3.6.) What would is 1. Use MATLAB’s “\” operation to solve the
you expect the rank of A to be? Explain. Use system Cx = b, and then compute the residual
MATLAB to check your answer. vector r = b − Ax.
(c) Let us now solve Cx = b by a new method that
(b) Which column vectors of A should form a basis takes advantage of the fact that A and C differ by
for its column space? Explain. If U is the re- a rank-1 matrix. This new procedure is called a
duced row echelon form of A, what would you rank-1 update method. Set
expect its first three columns to be? Explain.
What would you expect its last three rows to be? z = M ∗ u, c = v  ∗ y,
Explain. Use MATLAB to verify your answers d = v  ∗ z, e = c/(1 + d)
by computing U. and then compute the solution x by
(c) Use MATLAB to construct another matrix x=y−e∗z
D = ( E EY ), where E is a random 7 × 5 mat-
Compute the residual vector b − Cx and com-
rix and Y is a random 5 × 2 matrix. What would
pare it with the residual vector in part (b). This
you expect the reduced row echelon form of D
new method may seem more complicated, but it
to be? Compute it with MATLAB. Show that, in
actually is much more computationally efficient.
general, if B is an m × n matrix of rank n and X
is an n×k matrix, the reduced row echelon form (d) To see why the rank-1 update method works, use
of ( B BX ) will have block structure MATLAB to compute and compare
⎧ ⎫ Cy and b + cu
( I X ) if m = n or ⎪ ⎩ I X⎪ ⎭ if m > n Prove that if all computations had been carried
O O
out in exact arithmetic, these two vectors would
4. (Rank-1 Updates of Linear Systems) be equal. Also, compute
(a) Set Cz and (1 + d)u
A = round(10 ∗ rand(7)) Prove that if all computations had been car-
ried out in exact arithmetic, these two vectors
b = round(10 ∗ rand(7, 1))
would be equal. Use these identities to prove
M = inv(A) that Cx = b. Assuming that A is nonsingu-
Use the matrix M to solve the system Ay = b lar, will the rank-1 update method always work?
for y. Under what conditions could it fail? Explain.

CHAPTER TEST A True or False

Answer each of the statements that follows as true or 5. If S and T are subspaces of a vector space V, then
false. In each case, explain or prove your answer. S ∩ T is a subspace of V.

1. If S is a subspace of a vector space V, then S is a 6. If x1 , x2 , . . . , xn are linearly independent, then they

vector space. span Rn .

2. R2 is a subspace of R4 . 7. If x1 , x2 , . . . , xn span a vector space V, then they are

linearly independent.
3. It is possible to find a pair of two-dimensional
subspaces S and T of R3 such that S ∩ T = {0}. 8. If x1 , x2 , . . . , xk are vectors in a vector space V and
4. If S and T are subspaces of a vector space V, then Span(x1 , x2 , . . . , xk ) = Span(x1 , x2 , . . . , xk−1 )
S ∪ T is a subspace of V. then x1 , x2 , . . . , xk are linearly dependent.
27th December 2014 M03_LEON2218_09_SE_C03 page 183

Chapter Three Exercises 183

9. If A is an m × n matrix, then A and AT have the Span(x1 , x2 , . . . , xk ), then the vectors x1 , x2 ,. . . , xk ,

same rank. xk+1 are linearly independent.
14. Let {u1 , u2 }, {v1 , v2 }, and {w1 , w2 }, be bases for
10. If A is an m × n matrix, then A and AT have the
R2 . If X is the transition matrix corresponding to
same nullity.
a change of basis from {u1 , u2 } to {v1 , v2 } and Y
11. If A is row equivalent to B, then A and B have the is the transition matrix corresponding to a change
same row space. of basis from {v1 , v2 } to {w1 , w2 }, then Z = XY is
the transition matrix corresponding to the change
12. If A is row equivalent to B, then A and B have the of basis from {u1 , u2 } to {w1 , w2 }.
same column space.
15. If A and B are n × n matrices that have the same
13. Let x1 , x2 ,. . . , xk be linearly independent vectors rank, then the rank of AB must equal the rank of
in Rn . If k < n and xk+1 is a vector that is not in BA.

CHAPTER TEST B
1. In R3 , let x1 and x2 be linearly independent vectors 7. Let A be a 5 × 3 matrix of rank 3.
and let x3 = 0 (the zero vector). Are x1 , x2 , and x3 (a) What is the dimension of N(A)? What is the
linearly independent? Prove your answer. dimension of the column space of A?
2. For each set that follows determine whether it is a (b) Do the column vectors of A span R5 ? Are

R . Prove
subspace of 2 the column vectors of A linearly independent?
⎧ ⎫your  answers.
 Explain your answers.
⎪ x1 ⎪
(a) S1 = x = ⎩ ⎭ x1 + x2 = 0
x2 (c) How many solutions will the linear system

⎧ ⎫  Ax = b have if b is in the column space of
(b) S2 = x = ⎪ ⎩ x1 ⎪⎭ x1 x2 = 0 A? Explain.
x2 8. Given the vectors
3. Let ⎧ ⎫ ⎧ ⎫
⎧ ⎫ ⎪
⎪ 1⎪⎪ ⎪
⎪1⎪ ⎪
⎪
⎪
1 3 1 3 4 ⎪
⎪ x1 = ⎪
⎪
⎪ 2⎪⎪
⎪ , x2 = ⎪⎪
⎪ 3⎪⎪,
⎪
⎪0 0 1 1 1 ⎪ ⎪ ⎩ ⎭ ⎩ ⎪ ⎭
A=⎪ ⎪
⎪0 0 2 2 2 ⎪
⎪
⎪ 2 3
⎪
⎩ ⎪
⎭
0 0 3 3 3 ⎧ ⎫ ⎧ ⎫
⎪
⎪ 1⎪⎪ ⎪
⎪ 1⎪⎪
⎪
x3 = ⎪ ⎪
⎪ ⎪
⎪ ⎪
(a) Find a basis for N(A) (the null space of A). ⎪
⎩ ⎪
5
⎭
, x 4 = ⎪2⎪
⎩ ⎪
⎭
What is the dimension of N(A)? 5 3
(b) Find a basis for the column space of A. What
(a) Are x1 , x2 , x3 , and x4 linearly independent in
is the rank of A?
R3 ? Explain.
4. How do the dimensions of the null space and
(b) Do x1 , x2 span R3 ? Explain.
column space of a matrix relate to the number of
lead and free variables in the reduced row echelon (c) Do x1 , x2 , x3 span R3 ? Are they linearly inde-
form of the matrix? Explain. pendent? Do they form a basis for R3 ? Explain.
5. Answer the following questions and, in each case, (d) Do x1 , x2 , x4 span R3 ? Are they linearly inde-
give geometric explanations of your answers: pendent? Do they form a basis for R3 ? Explain
or prove your answers.
(a) Is it possible to have a pair of one-dimensional
subspaces U1 and U2 of R3 such that U1 ∩U2 = 9. Let x1 , x2 , and x3 be linearly independent vectors in
{0}? R4 and let A be a nonsingular 4 × 4 matrix. Prove
that if
(b) Is it possible to have a pair of two-dimensional
subspaces V1 and V2 of R3 such that V1 ∩ V2 = y1 = Ax1 , y2 = Ax2 , y3 = Ax3
{0}? then y1 , y2 , and y3 are linearly independent.
6. Let S be the set of all symmetric 2 × 2 matrices 10. Let A be a 4 × 5 matrix with linearly independent
with real entries. column vectors a1 , a2 , and a3 and whose remaining
(a) Show that S is a subspace of R2×2 . column vectors satisfy
(b) Find a basis for S. a4 = a1 − 5a3 , a5 = 3a1 − 2a2 + 4a3
27th December 2014 M03_LEON2218_09_SE_C03 page 184

184 Chapter 3 Vector Spaces

(a) What is the dimension of N(A)? Explain. (a) Determine the transition matrix correspond-
(b) Determine the reduced row echelon form of A. ing to a change of basis from the standard
11. Let {u1 , u2 } and {v1 , v2 } be ordered bases for R2 , basis {e1 , e2 } to the ordered basis {u1 , u2 }. Use
where this transition matrix to find the coordinates of
⎧ ⎫ ⎧ ⎫ x = (1, −1)T with respect to {u1 , u2 }.
⎩2⎪
u1 = ⎪ ⎭ , u2 = ⎪⎩3⎪ ⎭
3 5 (b) Determine the transition matrix corresponding
to a change of basis from the ordered basis
and
⎧ ⎫ ⎧ ⎫ {v1 , v2 } to the ordered basis {u1 , u2 }. Use this
v1 = ⎪
⎩ ⎪ ⎭ , v2 = ⎪
⎩ ⎪
4 2 transition matrix to find the coordinates of
⎭
7 6 z = 3v1 − 4v2 with respect to {u1 , u2 }.