You are on page 1of 7

Practice problems and concept list  not to be handed in

AMATH 352 Midterm Review

(1) Vocabulary
(a) Linear combination
A linear combination of two vectors is the sum of two vectors multiplied by
corresponding scalar values. Each term is multiplied by a constant that applies
to the entire vector.
(b) Linear dependence
A set of vectors is said to be linearly dependent if one of the vectors can be
dened as a linear combination of the others.
(c) Linear independence
A set of vectors is said to be linearly independent if none of the vectors are
linear combinations of each other. Additionally, in solving the system Ax = 0,
a linearly independent set will only have x=0 as a possible solution, meaning
there is no solution but the trivial solution.
(d) Subspace (of a `)
A subspace is dened as a euclidean space which is contained wholly in another
space. The characteristics of a subspace include: Having the zero-element,
being closed under addition, and being closed under scalar multiplication. We
say a space U is closed under addition if the vector a + b is U given that
aU and b U . We say that a space U is closed under scalar multiplication
if, given that a is a vector in U , a is also a vector in U for any R
(e) Span (of a set of vectors)
The "span" of a set of vectors is dened as the subspace occupied by all of the
linear combinations of vectors within the specied set.
(f ) Basis (of a subspace)
A "basis" of a subspace is a set of linearly independent vectors which have the
same span of the original subspace.
(g) Inner Product, Dot Product We dene the inner product of vectors x,y as the
sum of the products of corresponding elements. Thus < x, y >= x1 y1 + x2 y2 +
...xn yn . This denition is synonymous with our calculation of the dot product,
notated as x y.
(h) Norms
Norms are functions which take vectors as input and return non-negative real
numbers as output. Norms can be considered a means to measure the length
or magnitude of a vector. Norms satisfy a few conditions:
If x is R then kxk 0.
If x = 0 then kxk = 0.
If kxk = 0 then x = 0.
If R then kxk = |kxk.
1
2

Lastly, if x and y are both R, then kx + yk kxk + kyk, which satises the
triangle inequality.
(i) Linear functions
For a function f :U V
If u1 and u2 are U , then f (u1 + u2 ) = f (u1 ) + f (u2 ).
If u U and R, then f (u) = f (u).
Additionally, note that f (0) = 0, by taking = 0.
(j) dimension
The number of basis elements needed to dene a space is called the dimension
of the space.
(k) orthogonal (vectors)
Two vectors are orthogonal if and only if the lines drawn from the origin to the
points with coordinates dened by x and y are perpendicular to each other.
Additionally, the dot product of orthogonal vectors is 0.
(l) orthonormal (vectors)
Orthonormal vectors are both orthogonal and normalized (usually to 1 in the
case of QR decomposition).
(m) rank (of a matrix)
The dimension of the range of A is called the rank of A. This can be written
as rank(A ) = dim(range(A)).
(n) range (of a matrix)
The range of A is the span of the columns of A. This is also called the column
space of A.
(o) nullspace (of a matrix)
The nullspace of a matrix is dened as the set of vectors x which, when multi-
plied by the matrix A equal 0. A
Additionally, we can dene the nullspace of
as the set of vectors x AT .
that are orthogonal to the range of
T
We make the claim that Range(A is orthogonal to Null(A). This means that
for:
y Range(AT ) means y = AT x.
z Null(A) means Az = 0.
T T T T
Thus y z = (A x) z = x Az = 0.
T T
This suggests that Range(A A) = Range(A )
(p) inverse (of a matrix)
We dene the inverse of a matrix A as A1 , which when multiplied by A
becomes the identity matrix I.
Note that this means that both AA1 = I and A1 A = I.
(q) elementary vectors e1 , . . . , e n in Rn
The standard basis consists of the "unit vectors", e1 , e2 , ... en , which form a
basis for the space Rn . Applying a function to the standard basis vectors will
produce a matrix which models the function.
(r) Echelon form and reduced Echelon form
We dene a matrix to be in row echelon form when it satises the following
3

conditions:
The rst non-zero element in each row, called a "pivot", is a 1.
Each pivot is in a column right of the previous row.

We dene a matrix to be in reduced row echelon form when it additionally


satises the following condition:
Every leading coecient is 1 and is the only non-zero entry in its column.

(2) Concepts and skills


(a) Solving linear systems of equations
We solve linear systems of equations by creating an augmented matrix by which
the right most column represents the sum of the remaining left-hand columns.
Left-hand columns represent variables. Upon setting the augmented matrix
to reduced row echelon form, we can set the solution of the system of linear
equations equal relative to the coecients modeled by the pivots.
(b) Equivalence of linear functions from Rn to Rm and matrices
A function from f: Rn Rm can be modeled by a matrix of dimension Rmn .
Recall that we consider linear functions to be functions which satisfy two con-
ditions:
f (x + y) = f (x) + f (y). Compatibility with addition.
f (x) = f (x). Compatibility with scalar multiplication.
(c) Matrix multiplication
If A Rmr andB Rrn , then we can dene the product C = AB, which
will be a matrix in R
mn . Note that this product is only dened if the number

of coumns in A is equal to the number of rows in B. We produce this matrix


by multiplying rows by columns i.e.

a11 a12  
a21 a22
A= , B = b11 b12 b13
a31 a32 b21 b22 b23
a41 a42
thus

c11 c12 c13
c21 c22 c23
C=
c31

c32 c33
c41 c42 c43
Where

c11 = a11 b11 + a12 b12


c12 = a11 b12 + a12 b22

c43 = a41 b13 + a42 b23
4

(d) Finding a basis for a subspace (of vectors or polynomials)


We can nd a basis for a subspace by identifying the set of linearly independent
elements which span the span of the subspace. One method of doing so is
modeling the subspace with a matrix and reducing the matrix to reduced row
echelon form. The pivot columns of the matrix in reduced row echelon form
correspond to the basis vectors in the original matrix i.e. If a reduced matrix
has pivots in columns 1, 4, and 7, then the span of the original matrix can be
modeled by its rst, fourth, and seventh columns.
(e) Finding a basis for the range/nullspace of a matrix
To nd the nullspace of a matrix, solve the system for the zero-element.
Recall that the dimension of the nullspace and the dimension of the rank of
a matrix must add up to the dimension of the euclidean space in which the
matrix resides i.e. For a matrix A Rn , rank(A)+ dim(A) = n.
The range of a matrix can be found by identifying the span of its columns.
(f ) Row operations as (invertible) matrix multiplication
For any legal row operation applied to matrices, there exists a matrix which,
when multiplied by the host matrix, models this row operation. Additionally
row operations are reversible, and thus for each matrix multiplication which
models a row operation there exists a matrix which is precisely the inverse, and
reverses the operation.
(g) Finding the inverse of a matrix
To nd the inverse of a matrix, form an augmented matrix by which the left-
hand side is the original matrix and the right-hand side is the Identity matrix.
As Gauss-Jordan elimination is applied to the original matrix, the inverted
row operations which occur on the Identity matrix. Thus, when the left-hand
matrix is fully reduced to the Identity matrix, the right-hand side is precisely
the inverse of the original matrix.
(h) Linear regression

Practice problems
(3) Take any norm k k. Are the following norms?
(a) f (x) = 2kxk Yes. This is the same as taking the norm of 2 x.
(b) f (x) = k xk Yes, the norm of x is equal to the norm of x by the denition
of a norm. Therefore, a function which takes the norm of x is the same as
nding the norm of x and thus is clearly a norm.
(c) f (x) = kxk2 No, Clearly this does not work with scalar multiplication, as
f (2x) 6= 2f (x).
(d) f (x) = kx 2k No, The zero-norm is a non-zero element.

(4) Which of the following are subspaces?


(a) Lines through the origin (in Rn ) Yes. The sum of two vectors which pass
through the origin is another vector, which still passes through the origin.
5

Additionally, applying scalar multiplication to a vector which runs through the


origin will produce another vector which still runs through the origin.
R3 .
 
(b) All vectors perpendicular to 1 2 3 in
Yes. To be perpendicular or orthogonal, the cross product of two vectors
R3
 
must be zero. Clearly the set of vectors perpendicular to 1 2 3 in is a
subspace, as the sum of any two vectors which satisfy these conditions, u and
R3
 
v will still have a dot product of 0 with 1 2 3 in due to the properties
of matrix addition. Additionally, scalar multiplication of dot product which is
equal to zero results in a dot product which is still equal to zero.
(c) All vectors y so that hy, e1 i = 2. No. Scaling some vector y by a real number
6= 1 will change the inner product, meaning that this set does not satisfy the
condition of compatibility with scalar multiplication.
(d) Span of e1 , e2 in Rn .
Spans are dened as subspaces.
(e) Set of solutions to x2 = 0 in R. The only solution to x2 = 0 is x = 0. Adding
2
this solution to itself i.e. 2x = 0 still holds true for when x = 0. Scaling x = 0
by any real value will produce the equation x = 0

f : R3 R2 with
(5) Suppose I give you a linear function

1   1   0  
1 3 1
f 0 = , f 1 = , f 0 = .
1 2 1
1 0 2
Find an explicit matrix A so that f (x) = Ax. Add The rst column to the second
1

0   0
4
column. v2 = 1 = Divide the third column by 2 0 = 2 Subtract

3 1
1 1
2
3 7

1 0
the third column from the rst and second 0 = 2. 1 = 2

1 3
0 0
2 2

1 0 0  
0 1 0 = 1.5 3.5 0.5
0.5 1.5 0.5
0 0 1

(6) Which of the following are subspaces of F(R, R)?


(a) Polynomials of degree exactly 2. No, this does not contain the zero-element
(b) Polynomials of degree less than or equal to 2. Yes.
(c) Even functions, f (x) = f (x). Yes
(d) Odd functions, f (x) = f (x). Yes
6

(e) Functions f (x) : f (0) = 0.


Linear functions of 0 always
equal 0, because f (x) = f (x) holds true for
= 0 R Therefore the sum of any two functions which set x to 0 will produce
a function which still sets x to 0, and scaling a zero-function will still produce
0 as output.
(f )
4 00
Polynomials in p P , with p (1) = 1.
4 3
We can think of all polynomials of degree 4 to be in the form ax + bx + cx +
2

dx + e. This means the second derivative of degree 4 polynomials is 12ax + 2

6bx + 2c. Based on the condition p(1) = 1, this means that 12a + 6b + 2c = 1.
Clearly, if we were to add two polynomials which would meet this condition
1 4 47 3 13 4 47 3
i.e
12 x and x4 6 x , the resulting polynomial 12 x 6 x no longer meets
the condition, as it's second derivative will no longer equal 1. Therefore this is
NOT a subspace.

(7) Let

rs
S = {x R3 | x = 2s , r, s R}.
r
Verify that S is a linear space; then nd a basis for S. What is the dimension of S?

(8) Consider the matrix



2 2 2
A = 0 1 2
1 2 3
(a) What is the rank of A? Give a basis for the range R(A).
(b) What is the dimension of Null(A)? Give a basis for the null space N (A).
(c) What is the dimension of Null(AT )? Give a basis for the range R(AT ).

(9) Find a basis for the following space:

S = {p P 4 , p(0) = 0, p0 (1) = 0}

 
(10) Find four linearly independent vectors that are all orthogonal to 1 3 3 2 1
in R5 .

(11) Suppose you have a particular x


that solves a linear system Ax = b.
(a) Show that + v also solves Ax = b, for any v Null(A).
x
(b) Show that if x
solves Ax = b, then you can write x = x
+w for some w
Null(A).
7

(c) Now use the previous two problems to write down all solutions to the system

  x1  
1 0 1 0 x2
= 3
0 1 0 0 x3 2
x4
.

(12) Let

1  
V1 = 2 , V2 = 1 1 1 .
3
Be careful to think of these as matrices. Compute the following.
(a) V2 V1
(b) V1 V2
(c) V2 V2T
(d) V2T V2

(13) Suppose you want to nd a solution to x4 + x3 + x2 + 1 = 2.


(a) Explain how you would use Newton's method to do this.
(b) Starting with the base point x = 1, compute the rst iterate of Newton's
method for the problem.
(c) Can you also use the secant method? What would be dierent? Which would
you use?

(14) Write down the vector



1
2
1

1
as 1 v1 + 2 v2 , where v1 = 1 and v2 is orthogonal to v1 .
0

(15) A projector (onto subspace U) is a matrix P satisfying two properties:


1. The range of P is U and
2. Pu = u for any u in U.
1
(a) Show that for any v, the matrix
kvk2
vv T is a projector onto the span of v.
(b) Show that P2 = P.

1 1
(c) Compute the projection of 2 onto the span of 1.
1 0

You might also like