21.notes - Linear Algebra

Academic Year 2020-21 Linear Algebra
Roberto Costas (E223)
roberto.costas@uah.es
Bibliography LAY D e linear Algebra and its apps 14thEd
other material BB www.rscosan.com docencia.htm
Exams PEI PETE PEI 1 31 1131113
3019 W 10 00 12 00 in room SA5B 22515 Ll t
5 10 M 10 00 12 00 ONLINE
7 10 W 10 00 12 00 in room SA5B
12 10 M FESTIVITY
14 10 W 10 00 12 00 in room SA5B
19 10 M 10 00 12 00 ONLINE
21 10 W 10 00 12 00 in room SA5B
26 10 M 10 00 12 00 ONLINE
28 10 W 10 00 12 00 in room SA5B
Matrices, Determinants and Linear Systems
September 23, 2020

Matrices
A matrix Am⇥n is an array of numbers in rows and columns

0 1
a11 a12 · · · a1n ! r1
B a21 a22 · · · a2n C ! r2
B C
B .. .. . . .. C ..
@ . . . . A .
am1 am2 · · · amn ! rm
# # #
c1 c2 · · · cn
We say that the dimension of A is m ⇥ n (we also could say m-by-n). If

m = n, we say that A is square, otherwise we say that it is rectangular.

Matrices
Main diagonal. If A is a square matrix of dimension n, the elements aii ,

i = 1, . . . , n form the main diagonal of the matrix; the sum of these
elements is called the trace of the matrix.
Tr A a an t Ann
Transpose of a matrix. it is the matrix obtained when interchanging
rows and columns.

Generalities on Matrices
42 Es
Row matrix, column matrix.
too
E eDiag 2.3.51
Homework
aij o.i
Diagonal matrix.
j 1 ai.i
ae Identity matrix I. too IIe a eIs FIE Iz
U µg r ai o Null matrix O.
Zoo
Uij o i jam Triangular matrices (upper, lower). L
Symmetric matrix, skew-symmetric matrix. 1019
ATea ai.i ai.ie Diagonal by blocks matrix.
i
aij Aji AT ATdiagonal
Thisimpliesthatthe
entries are0
A y.bgso where
f t
Matrices
Operations: Khan Academy

1. Addition. Properties:
Commutative: A + B = B + A
Associative: A + (B + C ) = (A + B) + C
Neutral element: null matrix. A + O = A
Inverse element: negative of a matrix. A + ( A) = O
2. Multiplication by a number. Properties:
· (A + B) = · A + · B,
( + µ) · A = · A + µ · A
· (µ · A) = ( · µ) · A
1 · A = A.
Sometimes we will write

In the
Matrices, Determinants and Linear Systems page5 I

Matrices
3. Multiplication of two matrices. Properties:

Not commutative, in general. i.e. A · B 6= B · A.
Associative: A · (B · C ) = (A · B) · C
Neutral element for square matrices: identity matrix. A · I = I · A = A
Inverse element for certain square matrices: inverse matrix.
(A · B)T = B T · AT .
Obs: The matrices A and B are said to commute if A · B = B · A.
A is said regular, invertible if it has inverse, otherwise is said singular .
Matrices, Determinants and Linear Systems page 6

Matrices
detAto
ifAB Ac ACB 0 0 BC
Y
1
Inverse of a matrix: given an square matrix A, A (its inverse) is the
matrix, if it exists, fulfilling
Homework If AB I and AC I
A·A 1
=A 1
·A=I then B c
Moreover
BA I
A 1 does not always exist. A characterization of its existence can be
Homework why must A be n by n
achieved sby using determinants, or the notion of rank. wesolvedthisin
(A 1 T
) = (AT ) 1
. class
1 1 1
(A · B) =B ·A
Two ways for computing it: determinants or Gauss-Jordan method
(see later).

Result Any matrix can be factorized
m
by n
as follows
A LU L mx m U Mxn
wher L is a lowertriangular matrix

with main diagonal entries equal to 1
and U is a upper triangular matrix
This is the so called LV decomposition
then A LU where
If A is n
by n
entries are
V is n n with diagonal
by
Um Then the
Un Un
det A Ui Unit Um
The Gauvin method let us to

compute L au U
it is non zero entry

Example a pivot diagonal
II
a CUZ
A Laa o a
so det Ae AnAzz Arazi Infact pageG

Determinants
In fact if A is uppertriangular deLCA U Un Um

Given a square matrix A, the determinant of A, denoted by |A|, or
det(A), is a number associated with A.
|A| is defined first for 2 ⇥ 2 matrices. For matrices of higher order,
in principle it can be computed by expanding along any row/column.
For example, if A is 3 ⇥ 3, then we can expand along the first row
(we might also choose any other row or column)
a11
a21
a12
a22
a13
a23 = a11 · A11 + a12 · A12 + a13 · A13
You cantry by LU
a31 a32 a33 decomposition
where Aij denotes the cofactor of the entry aij (i.e. the signed minor
of aij ). Also, for 3 ⇥ 3 matrices, Sarrus’ rule may be useful.
Matrices, Determinants and Linear Systems page o

Determinants
Basic properties:
1. |A| = |AT |
2. If A and B are square matrices of the same order, then
|A · B| = |A| · |B|.
3. If all the elements in a row (or column) admit a same factor, then
that number can be taken out of the determinant.
4. If we interchange two rows (or columns), the determinant changes
sign.
Matrices, Determinants and Linear Systems page1

Determinants
Basic properties:
5. If A has a row or a column of 0’s, then det(A) = 0.
6. If A has two rows (or columns) which are either equal or
proportional, then det(A) = 0. The value is also 0 if there is some
row (column) which is a linear combination of others.
7. The value of the determinant does not change if we add to a row (or
column) other rows (or columns) multiplied by numbers. This
property is essential for efficiently computing determinants.

Determinants
Computation of the inverse of an square matrix A.

The inverse A 1 exists if and only if |A| = 6 0.
1 1
A 1= · AdjT (A) = · Adj(AT ), where Adj(A) is the cofactor
|A| |A|
matrix, i.e. for each i, j, the corresponding element is the cofactor
of aij .
we willexplainduringthenext
Alternative: Gauss’ method.
class
Matrices, Determinants and Linear Systems 1

Rank of a Matrix
We say that a certain row r (similarly for columns) is a linear

combination of the rows ri1 , . . . , ris if r can be obtained from these
rows by means of an expression like
↵ 1 · r i1 + · · · + ↵ s · r is
for certain numbers ↵1 , . . . , ↵s , which are called the coefficients of

the linear combination.
We say that certain rows (similarly for columns) are linearly
independent, if none of them can be obtained as a linear
combination of the rest. Otherwise, we say that they are linearly
dependent.
Question: When are two rows (resp. two columns) linearly dependent?

Rank of a Matrix
Definition
The rank of a matrix A, rank(A), is the maximum number of rows (or
columns) which are linearly independent.
An equivalent definition of rank, in terms of determinants. A minor,

in a matrix A, is any determinant that you can get by eliminating some
rows and/or columns. Then one may see that rank(A) is the maximum
order of the non-zero minors of A.

Rank of a Matrix
Some observations/properties:
We say that a square matrix of orden n has full rank (or is regular),
if rank(A) = n. It can be proven that this happens if and only if the
determinant of A is di↵erent from 0 (therefore, if and only if A is
invertible). If A is square and has not full rank, it is called singular;
such a matrix has no inverse.
The rank by rows coincides with the rank by columns.
rank(A) = rank(AT ).
If the dimension of A is m ⇥ n, then rank(A)  min(m, n).
When we compute the rank, we find rows/columns which are linearly
independent!
Matrices, Determinants and Linear Systems page 1

Rank of a Matrix
Some rules for computing rank(A):

A matrix has rank 0 if and only if all its elements are 0, i.e., A = O.
A row/column of 0’s does not count for determining the rank of A.
Similarly, a row/colum which is clearly a multiple of another
row/colum, or a linear combination of rows/columns, does not
count, either.
The rank does not change if we perform elementary operations on A
(swapping rows/columns, multiplying a row/column by a number,
add a linear combination of rows/column to another row/colum).
From a practical point of view, the computation of the rank can be
done either using determinants or by means of Gauss method.
Matrices, Determinants and Linear Systems page

Linear Systems: Definitions
A System of Linear Equations is a set of equations of the type

8
>
> a11 · x1 + a12 · x2 + · · · a1n · xn = b1
>
< a21 · x1 + a22 · x2 + · · · a2n · xn = b2
.. .. ..
>
> . . .
>
:
am1 · x1 + am2 · x2 + · · · amn · xn = bm
xi ’s: unknowns
aij ’s: coefficients
bj ’s: constant terms

Linear Systems: Definitions
The system can be written in matrix form in the following way:

0 1 0 1 0 1
a11 a12 · · · a1n x1 b1
B a21 a22 · · · a2n C B x2 C B b2 C
B C B C B C
B .. .. .. .. C · B .. C = B .. C
@ . . . . A @ . A @ . A
am1 am2 · · · amn xn bm
Abbreviately,
A · ~x = ~b
A: Coefficients matrix.
~x : vector of unknowns
~b: vector of constant terms.
Question: why is the above equality true?

Classification of Linear Systems
Classification of Linear Systems: linear systems can be

1 Inconsistent, if it has no solution.
2 Consistent, if it has some solution (i.e. it is solvable). In this case, it
can have
A Unique solution, or
Infinitely many solutions.

In order to classify a given linear system, we use the augmented matrix,

and Rouche-Fröbenius Theorem. The augmented matrix, that we
denote by B, is
0 1
a11 a12 · · · a1n b1
B a21 a22 · · · a2n b2 C
B C
B=B . .. .. .. .. C
@ .. . . . . A
am1 am2 · · · amn bm

Theorem (Rouche-Fröbenius Theorem)

Let A · ~x = ~b be a linear system of m equations with n unknowns, and let
B denote the augmented matrix of the system. Then the system is
consistent if and only if rank(A) = rank(B); furthermore, the system
has a unique solution if rank(A) = rank(B) = n, and infinitely many
solutions if rank(A) = rank(B) < n.
When rank(A) = rank(B) = n, the di↵erence n rank(A) is in fact the

number of degrees of freedom of the system, i.e. the number of
parameters the solutions depend on.

Solving Linear Systems
Two possibilities:
1 Cramer’s Method: uses determinants and must be applied on a
Cramer’s system (i.e. a system where the coefficients matrix has full
rank). It is not efficient for big systems.
2 Gauss and Gauss-Jordan Method: does not require to compute
determinants, but just simple operations with rows/columns.
Efficient for big systems.

Homogeneous Linear Systems
These are those linear systems where the constant terms are all 0:
8
>
> a11 · x1 + a12 · x2 + · · · a1n · xn = 0
>
< a21 · x1 + a22 · x2 + · · · a2n · xn = 0
.. .. ..
>
> . . .
>
:
am1 · x1 + am2 · x2 + · · · amn · xn = 0
Always consistent (why?)

The interesting question is whether it has other solutions, up to the
trivial one (in such a case it has infinitely many!)
This happens only if rank(A) < n.
If A is square, this is equivalent to |A| = 0.

A 2 3 compute the determinant
of A 1 if P
A 2 A 4 1 O O Ui Ur UB Z
O
A LU l T O O Un Uu
lE l3 r 0 0 UE 2
3 b
Azz 2 L Ucz l2Uktl3Urz 313 13 3

Azz I Vcs lz Uist13h23 t Uz3 34 1 433 U33 31
1
there def A UnUnUss L I 371 137
1
By Samus's method 1 0 12 O 8 4
A
adjugate matrix of
F
The AdjCA
T
AdjIADi.j C hitidetfAlpj.cl
matrix A where I throw
and
i thcolour is
deleted
Examples
da ab
If A Ibd ADELA
at c t.aaitaif.IE1
I ititl page2
Remember that there are di erent ways to solve a linear system. For example by using the inverse
of the matrix of the coe cients:
II
AI 3 A AI Ab e.at x A b
1
Let us compute A In two di erent ways.
1) By using the formula for the inverse:
A 2 3 7 4 5
A l
A 2 A 4 a r z
O 2 A
1 4 2 3 inver
ofA
2) Use the Gauss elimination method: The idea is to go from A Il n I B
Pwt
pivot pivot
yo o 125 3123
A 2 3 I A 0 O RT Rz 212 a 2 3l A 0 0
2 A 4 I O t O O 3 21 21 O
RT RT12122
I
O 21 10 01 O 21 10 01
i
wt pint
P'Io2 3 z o g
l l 2
O Ri Ri 15123
A A 0 O RT 3121 I
RT Rz 2123
O 3 21 210 RT RT 12122 O 3 2 210
i
O O l 1i 423 o o l 1i 423
At
5 43
3 O O I 211215 12 I 0 O 745
122 3
O 3016 3 6 O l O I
212
0 O l 1i 423 R Rzf O 0 A l14 2 3
page25
We also solved problem 11 of sheet 1
A a compute W decomposition
A Ja
As
b by using such decomposition
solve A I 11,3DT
3
a
f a r r
µ It
from here we have working
A LU Oo03
3
2 with the rows
3
l
Are re rn Ai l l 6e 3
6 e 3
Az lire r2 s 13,6 6 h h l 0 T
Az her G r2 t B s
2,1113 la la le 0,313,313 t QQ
lz 11 12 313 le
2 13 k 34
f if ya
Bo Egg
b AI Lu
then
f HH I
vii i
fix HH til
page26
If we considerthe
Euclideanplane 1122
Lesson 2: Vector Spaces n r 7 2,0 1

g
slope
n the same space R2 vectors

September 15, 2019
l
µ Belo
k
s
1
numbers coexist
points and complex
Raff o
r A two pointsof
r
Lesson 2: Vector Spaces
Vector Spaces
Let V be a set, and let +, · be two operations, the first one (sum)
defined between the elements of V , and the second one (product by
scalars) defined between V , and the elements of R (resp. C). We say
that (V , +, ·) is a vector space over R (resp. C) if the following
properties hold:
(i) (V , +) is a conmutative group:

Internal Law: 8~ u, ~v 2 V , u~ + ~
v 2 V.
Associative: 8~ u, ~ ~ 2 V , u~ + (~
v, w v +w ~ ) = (~
u+~ ~.
v) + w
Neutral element: there exists ~0 2 V such that u~ + ~0 = u~.
Inverse element: for all u~ 2 V , there exists u~ 2 V such that
u~ + ( u~) = ~0.
Conmutative: 8~ u, ~v 2 V , u~ + ~
v =~v + u~.
Lesson 2: Vector Spaces page

Vector Spaces
Let V be a set, and let +, · be two operations, the first one (sum)
defined between the elements of V , and the second one (product by
scalars) defined between V , and the elements of R (resp. C). We say
that (V , +, ·) is a vector space over R (resp. C) if the following
properties hold:
(ii) The operation · satisfies that:

8 2 R (resp. C), 8~ v 2 V , · (~
u, ~ u+~v ) = u~ + ~ v.
8 , µ 2 R (resp. C), 8~ v 2 V , ( + µ) · u~ = u~ + ~
u, ~ v.
8 , µ 2 R (resp. C), 8~ v 2 V , · (µ~
u, ~ u ) = ( · µ) · u~.
1 · u~ = u~.
Lesson 2: Vector Spaces page2

Vector Spaces
In order to make explicit whether we work over R or C, one writes

(V (R), +, ·) or (V (C), +, ·). In the sequel, we will assume that we work
over R, although the results are equivalent for C. In fact, vector spaces
can be defined over sets other than R or C (fields).
Inthe previous example the space of all

Examples
uultip
af Fr is a rectorspace
A E Maz IR A AT is a vector space youmustcheckthat
2 2 Symmetric matrices with real entries
by
5k a bicep R W
I thasdimension 3
then dimensional
f V W visas
f abba Carb c for example
Two elements of V are
ii
Is hit ME
at Ma
Hottub
db µb
da pic
E 1 ab
page 3
Vector Spaces
Proposition
Let (V (R), +, ·) be a vector space over R. Then for all 2 R and for all
u~ 2 V , it holds that:
(1) · ~0 = ~0.
(2) 0 · u~ = ~0.
(3) · u~ = ~0 if and only if = 0 or u~ = ~0.
(4) · ( u~) = ( ) · u~ = ( · u~).
Lesson 2: Vector Spaces page 3

Linear Dependence. Bases.
Definition
A linear combination of vectors u~1 , u~2 , . . . , u~n is another vector of the
form
~1 + 2 u~2 + · · · + n u~n
1u
where i 2 R for i = 1, . . . , n.
Remark The vector ~0 can be considered as a linear combination of any

vectors u~1 , u~2 , . . . , u~n , because ~0 = 0 · u~1 + 0 · u~2 + · · · + 0 · u~n .

Definition
We say that {~u1 , u~2 , . . . , u~n } are linearly dependent (l.d.) if at least of
them is a linear combination of the rest. If {~ u1 , u~2 , . . . , u~n } are not l.d.,
we say that they are linearly independent (l.i.).
3
Lesson 2: Vector Spaces 3L
1
A
Kiffin tix l
4
l m
we multiply w by rows and we compare both expressions

row
row by
Aa Ui U 1,2
Az like Ur 4,91 14,241 101
9 2 lit 9 8 1
ae kitsune i
l X LID
king F F
Ez a un p
vesta foil stil I H L
3 2 2 37
page 3
At
2 A L LDLT Ac Ui Ui Cli
Ar 1427
Az like
4
It 22
A
0 2,47 4,247 197
day 0 uz
ou 7 4 24 0
119361811
T
1
L D L
page36
3 112
U
a
III III to I U
In 2 4 3 2
Az l U a Uz Cl Z l 24 li le t o
Az lzUielzUztUz 1 24 I
h 1 4 32
Cy 1 2 2K ez le o
Elz 3b poiQ
T
d
2h2 I lz 2 lz1213 2232 0
l ez 13
213 Z lz Z l3
Homework i compute D so that
A LDF
page37
i
4 a
f EH Qz l canal
o o 17
li 24 24 16
5124 9
A z e l U t Uz 2,7 5 y
A3 KU el3UztUs 24 3
y
I
C 45 307 ez Zee 2 la 0 3h3 9 t o Q
T t
lz
f
30 26 913
5 2h2 1313 l3 30 4 27 I
aint.LI 1 9.11
I
page38
5
Efg
A 14g LU solve AI's
qs
a
b
1
0
3
B iioioii.si
lo
610,011
A2 44 42 14 li le t
i 2,347 YK 6 213 1
lzU tl3UztU3 le lzilz.la

4
A t Ql3i2l3,3l3 0,0
T T
T L
1,316,10 10 12 313 3
3t3 lz
2 Ay.lqUitlgUztl6UztU4
llqlu.lu.la tf0pils 2lsn3ls t
0,01 36 610,0
114,10120 r t d
Zo ly Bls 36 1
Is 4 ly 3 Is lo la 2ls 3
UE JLF ff.gg sUI yf
page 3
6 Homework
112
7
f
a.t Hof Il O
l
Eteria
0,4 Co
33 27 144 24,0 0 l
t b
2 Z
o 2,6 Gee 2 ez co o 2ez 2 G Cola
i 2h3 2 ls 1213 4
a.t
A is PositiveDefiniteIPD if I A IT o for E 1 0
If A AT A P D when A _LDLT D has positive

diagonal entries
A Ll U LCD LT A is
if A LU e
negative
LD LT definite
symmetric
page40
a
III II.it0981 E
o
t fco.zn
O Co
213
o 4
3
c in 17 124 lied O
2 4 2
374g
li I I
f
01 112 Gee le de Co Zf G 019
l ltlz zlz ls
IE II Et
z a.
Inverse
uor detA
2 zZx z
of A by using
4 to A is regular
Gauss Jordan method
ft O O Ri R2
y 2 f l o e O T2z Rzt2R
A'III I 2 Ii or o a a oil 00
O l 2 I 0O l O l 2 1 00 A
pivot i
e
l 2 l O l O p
f Rz I 2
l O l O
l RI Rzt3R2
O 3 2,1 20 O l 2 1 00 A
O l 21 O O l
i
O 3 211
l 20
l 2 I O l O E O 3 101 2
RT Ri ZB
O e 2 OO l
k
f e 2 OO l
Y
k 2 tzR3
O O 4 11
I
23 O O 4 1 23
y o o i I E E RI R
y o Z EE
o l
o e O tI i z o e O l I i E III A
4
O O 4 11 23 RT 4123 O O 1 1I
I
page41
For vectors in Rn : in order to check the linear independence of

several vectors, form a matrix with them as columns (or rows), and
compute the rank (Example)
The above idea can be extended to other cases (but we need the
notion of coordinates of a vector, to do this; we will see it later).

Theorem
The vectors {~ u1 , u~2 , . . . , u~n } are linearly independent if and only the only
linear combination of them fulfilling 1 u~1 + · · · + u~n = ~0 satisfies
1 = · · · = n = 0.
page43
Definition
We say that S = {~ u1 , . . . , u~n } is a spanning set of V if any vector in V
can be written as a linear combination of the vectors in S.
Definition
We say that B = {~ u1 , . . . , u~n } is a basis of V if it is a spanning set of V
and they are linearly independent.
Examples
Important remark: vector spaces may have infinitely many bases!!

page4L
Definition
We say that a vector space has finite dimension if it has a basis
consisting of finitely many vectors.
Examples:
1 Rn has dimension n, since it admits the basis (called the canonical
basis)
{(1, 0, 0, . . . , 0), (0, 1, 0, . . . , 0), . . . , (0, 0, 0, . . . , 1)}
Notation: e~1 = (1, 0, . . . , 0), e~2 = (0, 1, . . . , 0), . . .,

e~n = (0, 0, . . . , 1).
2 Matrices of fixed dimension, polynomials of fixed degree.
3 At least an example of a vector space which is NOT this kind?

page45
Theorem
If V has finite dimension, then all the bases of V have the same number
of vectors (the dimension of V , dim(V )).
Intuitively, the dimension of a vector space is the number of parameters

that one needs to specify in order to identify a concrete element in the
space.

page46
Theorem
Let V be a vector space with finite dimension, and let dim(V ) = n. Then
the following statements are true:
(i) If S spans V , then you can extract a basis from S.
(ii) Every system consisting of more than n vectors is linearly dependent.
(iii) Every spanning system contains at least n vectors.
(iv) Given B = {~ u1 , . . . , u~n } ⇢ V (a subset of exactly n vectors), the
following statements are equivalent: (a) B is a basis; (b) B is linearly
independent; (iii) B is a spanning system.

page47
Definition
Let B = {~ u1 , . . . , u~n } be a basis of V , and let ~v 2 V . The coordinates
of ~v with respect to B are the scalars 1 , . . . , n 2 R such that
~v = ~1
1u + ... + ~n
nu
Usually we write
~v = ( 1, . . . , n )B ; we also say ~v has coordinates ( 1, . . . , n) in B.

page48
Theorem
Let V be a vector space of finite dimension n, and let B = {~
u1 , . . . , u~n }
be a basis of V . Then every vector ~v 2 V has unique coordinates with
respect to B.

Vector Subspaces.
Definition
Let (V (R), +, ·) be a vector space. We say that W ⇢ V is a vector
subspace of V if (W (R), +, ·) has also a structure of vector space.

Vector Subspaces.
Theorem
W ⇢ V is a vector subspace if and only if 8 u~, ~v 2 W , 8 , µ 2 R,
u~ + µ~v 2 W (i.e. if and only if every linear combination of two vectors
in W , stays in W ).
Examples
Observation: If W is a vector subspace, then ~0 2 W .

page51
Vector Subspaces.
Since a vector subspace is in fact a vector space inside of another

bigger vector space, it makes sense to speak about its dimension,
about bases, etc.
Again, the dimension of a vector subspace is the number of
parameters defining an element of the subspace.

Vector Subspaces.
Definition
u1 , . . . , u~n } be a subset of V . The linear variety spanned by S
let S = {~
(or simply the linear span of S) is the set consisting of all the vectors
which are linear combinations of the vectors in S, i.e.
L(S) = {~x 2 V |~x = ~1

1u + ··· + ~n }
nu
L(S) is a vector subspace, and its dimension is the rank of the

system S, i.e. the number of linearly independent vectors in S.
In R2 , the linear span of a vector is a line.
In R3 , the linear span of one vector is also a line; the linear span of
two independent vectors is a plane.
In Rn , n 4, one vector spans a line, two independent vectors span
a plane, three independent vectors span a space,...

Vector Subspaces.
Equations of a vector subspace:

1 Vector equation.
2 Parametric equations.
3 Implicit equations.
Change of basis

page54
SHEET 2
1. (c). Is NOT vector space because

plt 0 Is not in the space. the polynomial identical
2 217107 p o
to Zero J
T T
condition because pltko
(a) Yes.
If E Cxy EW ELITE and
and 2e IR
and I cxiyiziewa.ie fX5LfyIE
U tT XtX yy 2 z EW
we need to check if it ful lls the equations:
0
ly 14 1212 170
2 0 0
Xtx't
3CXtx 114 4 01 0
tu Ew we need to check if it ful lls the equations:
0
Ix Hy 1222 40
3147 ay 40 0
122 0
Obtaining a basis: X Y implicit equations
We O
3 1 y
pint
It has in nitely many solutions
Iii L
page55
you must use 1 parameter
a 121 4
x
1 12 X y 24
E Rz 3121
O 4 6 g 4g 61 y 21 34
n
2 D z 21 Dim W 1
WeSpace C l 3.2 5 1
M Mt W
b ME1433 IR letM 1 few and Mz bodied
few andDEIR
M 1 Mz CW ata btb Ctc
Mi M2 btb dtd ete EN
j µ EW etc ere ftf
AM EW It is analogous
Implicit equations: X
X6
4
X Xz X3 And
X7 X3
Xs Xt
µ tew xu
g
Xa Xs 19 Xq
X Xs
pink free parameters
to'i
G OO OO 1 O 1
III parameters diucw
So the paramehiceq.ua are
X H 1 116
2 62 Xp 14
z by
and
g y
Xy dz Xa 46
Xs d3
page56
In order to compute a basis, you set each parameter equal to 1, and the rest equal to 0 to get every
element of the basis.
BA ofI
del Iz 26 0
Melo Ei
de0,22 1 HE 76 0
wept Isomorphism
Nzef EztE4
he
disco too as 7,74 96 0
Carb c die f
h E 00
3 1868 Es
6,08888
O 00 I 00 1
4 a so.arnasaoo
M4s do EstEz
die 1450 25 1 CIG 0
Ms180,8019 Estes
d 95 0,96 1
Me
1 f Eg
5
page57
9. Let us consider the matrices
0 1 0 1
1 0 0 1 1 1
A = @2 1 3A B = @1 2 2 A
0 0 1 1 2 3
Obtain, by using the Gauss-Jordan method, the inverse of A, B and AB. Check that
(AB) 1 = B 1 A 1 .
Al
a naii.at lie i'I

as me li i
HiEEiatooioI
i tE EL iiog
o Il
B
I I
fempu
I 00
f
AB In II ABY GS method
AB fool
o
EEE
ki B t I EEEur.fi 1
fEedEEEiII l
i iii'iit
2 I O
it
l O0
AB
BA
it I 9 I3
2 I 4 1
page58
10. Let A be the 2-by-2 matrix: ✓ ◆
1 2
A=
3 7
By using the Gauss-Jordan method, compute the inverse matrix of A and of its transpose
AT and check that (AT ) 1 = (A 1 )T .
at I 3I I si I is
t's I I II is
a
Iii
10
page59
11. Let A be the matrix 0 1
1 1 1
A = @3 6 6 A
2 11 13
(a) Obtain its LU decomposition.

(b) Solve the linear system Ax = (1, 3, 0)T by using the decomposition obtained in (a).
Homework
a
at L
It U
b Ux y i Ly L y ft LI
Ux y
11
page60
12. Let B the matrix 0 1
1 1 1
B = @3 6 6 A
2 11 k
(a) Obtain the values for k so that B is singular.

(b) Find the values for k so that the linear system Bx = (2, 9, 13)T has infinitely many
solutions. Obtain the general solution for such a case.
(c) Find the values for k so that the linear system Bx = (2, 9, 13)T has one solution.
Obtain such solution.
LU decomposition
F Uk
13 detcpg 1 3 k 117 0 if kid
13 11293,81181 Bissinger
n t
L KM
thx y ly111 g
a
X Xr 1 3 2 2 X3
31 2 31 3 31 z
we set Kitt 0 0 free Ea X htt

2g
uax.yly f.IT I
I 2 Diez O l D
Unity I k µ
i E to
O
12
page61
13. Let 0 1
1 1 2 0 0
A= @ 1 2 2 2 4A
1 0 2 2 4
(a) Solve the homogeneous system Ax = 0.

(b) Solve the underdetermined linear system Ax = (2, 4, 0)T .
(c) Given three values, namely b1 , b2 , and b3 . Find the linear system such values need
to fulfill so that the linear system Ax = (b1 , b2 , b3 )T has solution.
n n
a
f II IE O l O 24 bzibis
EfEifii ai
yE
Iiio OO OO og
bist2b bz
Rank A 2
Discuss 2 if 2b bz o
RankA
3 if by124 bz1 0
When rank A rankCA't
When do we have a solution
b bz bz o there is always a solution

Rewind When
13
page62
LINEAR ALGEBRA (350000) Course 2019/20
Practice exercises .. Sheet 2
Vector Spaces. Change of basis
1. Analyze whether the following subsets are vector subspaces of the given ones. In affirma-
tive case, obtain a basis for such subset.
Homgen Lin
Sys
(a) W = {(x, y, z) 2 R | x y + 2z = 3x + y = 0} of R
3
YES Dim 1
3
T
(b) W = {M 2 M3⇥3 (R) | M = M } of M3⇥3 (R) YES
Dim 6
(c) W = {p(t) 2 P2 (R) | 2p(0) p(1) = 2} of P2 (R) No
IT Is not homogen
1123
142 17,191,17 1,01 3 Basis of
4411 0 it is not a basis
2. Analyze if
(a) B = {(1, 2, 1), (0, 1, 1), (1, 0, 3)} is a basis of R3

⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 1 0 0 1 0 1
(b) B = , , , is a basis of M2⇥2 (R)
0 0 1 1 1 0 1 1
(c) B = { x2 + x + 5, 3x2 + 5x + 3, 2x2 + 2x 1} is a basis of P2 (R)
is viii ciii Basis as Mazur

We consider two ways:
D1 dzVz bB 1yFy E Kidzidsdy O

1
if YES Lin ind
if NO Lin def
1 page63
I
1
H 11 11
In
H O l O f
lO t.li H ii i J
2 MazCR 1124
t Ebd
Is largebicypd
the coordinate matrix is
O l O 1
p B
RIN RIN BIH
c
f XIX
2
15,3 5 3 2
2 2 1 Basis of PzCIRI
1124127 e IRB with flax't b Xt c Ca b o
f
of
a
1123 PsPups ki in 112412 is
are
f pix l L ti 5 c
equivalent to prove that the
rector
are E
f l P2 3,5 3 i in pe
HI Pa 3 are l
f Ps 2 z e c R3
irs
page64
They do notform a basis
det 353
1 5 30 6 50 31 6 0
Change of Bases Br thx 13 1,17 Con

Bz l l X oil Hit
fi 2
Pdx linear mappings
IRCA Pdx
g Bc 132 A
By il
xd
g fzof
p T T
Az Ai coordinates
B
B 2
Here rtx a I Bu 1 I'tB f
T
Atx Hi B Bz 21 1 Be coordinates
B
and a
t
x a Bu o I p
1 Xi Aco Pz
Atx 1107131 canonical

Observe that i
0 1 B1 basis 6
1 page65
Therefore i
Dnt Matrix ofthe
2 1
g tho Kil ga B then B y o
change of
Bases
g 0.14 11,0
B
1 O is it
f As As y 1 MpaBc B AzA
fzs Az Az Oy
1 fly to MBe.BZ
Gauss Jordan mitin
gEliit I
O
msn.im I is
page66
Lesson 3: Linear Mappings
September 15, 2019

Linear Mappings

Linear Mappings
f injective

Linear Mappings
f surjective

Linear Mappings
f bijective

Linear Mappings
f bijective

Linear Mappings
Definition
We say that a mapping is injective, if there are not two di↵erent
elements of S with the same image. Also, we say that a mapping is
surjective if every element of S 0 is the image of some element of S.
Finally, we say that a mapping is bijective if it is both injective and
surjective.
Here we will be interested in a particular kind of mappings, namely linear

mappings.

Linear Mappings
Proposition
Let f : V ! V 0 be a linear mapping. Then it holds that:
(1) f (~0) = ~0.
(2) If {~
u1 , . . . , u~n } are linearly dependent, then {f (~
u1 ), . . . , f (~
un )} are
also linearly dependent.
(3) If S ⇢ V is a vector subspace of V , then f (S) is a vector subspace
of V 0 .

Matrix Equation of a Linear Mapping
Let f : V ! V 0 be a linear mapping between vector spaces, and let

B = {~u1 , . . . , u~n } be a basis of V , and B 0 = {~
u10 , . . . , u~m
0
} be a basis of
0
V . All that we need in order to define f , is to know the images of
the vectors in B.
Matrix equation of a linear mapping. Examples
Lesson 3: Linear Mappings stop

Properties:
(1) Given f : Rn ! R, f is linear if and only if
f (~x ) = f (x1 , . . . , xn ) = a1 x1 + · · · + an xn
Similarly when f : Vn ! V10 (i.e. if the final space has dimension 1).
(2) Given f : Rn ! Rm , we can write
f (~x ) = (f1 (~x ), . . . , fm (~x ))
where each fi is said to be a component of f . Then, f is linear if

and only if each fi (~x ) (which is a mapping from Rn to R) is linear
(see (1)). Similarly when f : Vn ! Vm0 (i.e. when f goes from a
space of dimension n, to another space of dimension m).

Properties:
(3) Every linear mapping can be written as
~y = A · ~x ,
where A is said to be the matrix associated with f (in certain

bases B and B 0 ). If f : Vn ! Vm0 , then A 2 Mm⇥n .
(4) The columns of A are the images of the vectors of B under f ,
expressed in the basis B 0 .
(5) The matrix A depends on the chosen bases B, B 0 : if we change
some basis, A changes too!

Properties:
(6) Let B, B 0 be fixed bases in V , V 0 respectively. Given a linear
mapping f : V ! V 0 there exists a matrix A associated with it in
the bases B, B 0 . Conversely, any matrix A defines a linear mapping
with respect to the considered bases. So,
Matrices $ Linear Mappings
Examples of linear mappings: Khan Academy (click)

Proposition
Let f : Vn ! Vm0 , g : Vn ! Vm0 be two linear mappings, with associated
matrices Af and Ag , respectively, and let k 2 R. Then, it holds that:
(1) f + g is also linear, and its associated matrix is Af + Ag .
(2) k · f is also linear, and its associated matrix is k · Af .

Proposition
Let f : Vn ! Vm0 , g : Vm0 ! Vp00 be two linear mappings, with associated
matrices Af and Ag , respectively. Then the composition g f : Vn ! Vp00
is also linear, and the matrix associated with g f is Ag · Af .
We can apply this to study how A changes when the bases B, B 0 are
changed. Khan Academy (click)

3. In the space of polynomials of degree, at most, 3 with real coefficients we consider the
vectorial subspace spanned by the polynomials fulfilling the conditions
BIR
p0 (1) = 0, p0 (2) = 0. 11
Space
x3x3xm
Obtain a basis of this vectorial space.
u
ax3tbxtcxdia.bgd
R
W pCx7EPz Rl p'll 0 p 2 so
ax3tbxtcx di3at2btc ogl2 at 4btC o
ERe412 c b
I pot's
free
a r
3
One Basis of W is 2 9 412X t
4 12
4. Obtain the null space and the vectorial space generated by the columns of the following
matrices: 0 1 0 1
1 2 0 1 1 2 1
B0 1 1 C B0 4 2 2C
A=B C B CRzRs R
@2 1 3 A B = @ 1 3 0 3A
1 0 0 0 1 1 1
coecahspanffql.fi d collB7 span Tg

II I
Null A xeR3 AX03 1 Null B xeR4iBx o
dim NAIA coeCA rank A

span
f
2
page68
5. Given the vectors v1 = (1, 2, 1, 1), v2 = (2, 3, 1, 2), v3 = (1, 3, 2, 1) and v4 = (2, 1, 1, 2)
of R4 , we want:
(a) To know if they form a basis of R4 .

(b) If not, we want to complete a basis of R4 and obtain the coordinates of the vector
v = (4, 1, 0, 1) in such a base.
a To check that
I To 30 0 Therank is 2
bl 1134 spa Iii riiio.atoico.oYiin3 Basis to
4,110111 2WitBWz18W fwy linearsystem
I 20 I 10
IZ
a
A
Iii
flo
L
2,00
O UX
tl H t ff 3 3
6. Given the the basis of R3

7
B1 = {(1, 0, 2), (0, 2, 1), (1, 1, 2)},
obtain the matrix of the change of basis from B1 to the canonical basis Bc and from Bc
to B1 , where the canonical basis of R3 is
Bc = {(1, 0, 0), (0, 1, 0), (0, 0, 1)}.
Given te vector u = (1, 1, 3). Obtain its coordinates with respect to the basis B1 .
Men Bc i MBcih Mi.BE

f FI
Ui 11 113 11 1,37g coordinates u withresp

MpaB
of to Bs
12,5 HIBS 3
Isn
7. We consider the following basis of R3 :
B1 = {(0, 2, 2), (2, 0, 1), (3, 0, 0)} and B2 = {(2, 0, 2), (1, 2, 0), (0, 0, 1)}.
(a) Obtain the matrix of the change of basis from B1 to B2 .

(b) Given the vector (1, 1, 1) with coordinates with respect to B1 , obtain its coordinates
with respect to B2 .
Miss.HN
eszMrs Bc fEIfJf
t.EE 1 b7U tii
C
llcl.Dpg 27BzMBnBz ti Ig
21h 3 Bz
4
8. In the vectorial space P1 (R) we consider the basis
B1 = {t, 1} and B2 = {1 + t, 1 t}.
We want to:
(a) Obtain the matrices of change of basis from B1 to B2 , and the one form B2 to B1 .
(b) Given the polynomial p(t) = t+3, use the correct matrix of change of basis to obtain
its coordinates with respect to the basis B2 .
Weassume Bc I413 13 t H Be 1 6117,3

3414are l i if Ilo fIto
a MBBZ Mais MBBo MI Ba MBBe L I df
t
MBZBi MI.BZ L
b plt te 3 13 L ME
B
C2cNB MeaBzL's
f zl
5
9. In the vectorial space M2⇥2 (R) we consider the basis
⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 0 1 1 0 0 1
B1 = , , , .
0 1 1 0 0 0 0 0
Find the matrix of change of basis from the canonical basis of M2⇥2 (R) to B1 , and obtain
the coordinates of ✓ ◆
3 4
A=
2 1
with respect to B1 .
1124 t by 1 correspondence
f MexCR t the
coordinates of Ibd w r
Edb s a b cd
can basis of Muz
m.am nisi
1 too O l l 0
A LY s 13.4.41 Be L Bs t 214 6
By
MBe.iq
J fIgJL
6
10. In the vectorial space M2⇥2 (R) we consider the canonical basis Bc , and the basis
⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 0 1 0 1 0 0
B1 = , , , .
0 0 0 0 1 0 1 1
We want to:
(a) Find the matrix of change of basis from B1 to Bc , and from Bc to B1 .

(b) Given the matrix ✓ ◆
1 2
C=
2 1
obtain its coordinates with respect to the basis B1 .
Homework check the solution in the PDF
7
11. Given the following basis of R4 :
B = {(1, 0, 2, 1), ( 1, 1, 0, 0), ( 1, 0, 2, 1), (1, 1, 0, 1)},

a
we want to obtain the matrix of the change of basis from B to the canonical basis of R4 ,
the matrix of the change of basis from the canonical basis of R4 to B, and the coordinates
of the vector (1, 2, 0, 1) with respect to the basis B. b
a p is a basis GI
11 41471 14291 4 the elements 413

are li
msn.LY Mimi
I liked
b E Hi 2917 1 p
MBE.B 92 gaff
pivots
Ite a n ooo RTR 212 D988Esr
Interchange
II
2,2 e a r ooo
o o iii or oo 12441 dh o o iii or oo Rz R4
20 20
r o r e l
OO t O
0001 f
z 4 2 2010
a 2 O 1001 If o 4 o 2 210
z r l i 101
RI Ry 212
425
Rs 1
r
10 00 1 o RT123 2124
l l l 1 O hO ReRT Re Re 124
2100joe
21 a
88 o
2 I
io
I
no.no
Er
iii't
Oi
O t O 0
10 It
s 1001
MBcB
8
12. In the vectorial space P3 (R) we consider the canonical basis Bc = {1, t, t2 , t3 } and the
basis B1 = {t2 + t3 , t + t2 , t, 1}.
We want to:
(a) Obtain the matrix of change of basis from B1 to Bc , and from Bc to B1 .

(b) Given the polynomial 1 + t + t2 + t3 , use the right matrix of change of basis to obtain
its coordinates in the basis B1 .
MB
L Mars nisi a
f
b htt Etf 4111,1 1 Be C 7 By 4011d Be
msn.li
fIl
9
13. In the vectorial space P3 (R) we consider the basis B1 = {t3 , t2 , t, 1} and the basis B1 =
{t3 , t + t2 , 1 + t, 1}.
We want to:
(a) Obtain the matrix of change of basis from B1 to B2 , and from B2 to B1 .

(b) Given the polynomial t3 + t2 + t, use the right matrix of change of basis to obtain
its coordinates in the basis B2 .
Check the solution in the PDF
10
5th
Week
Definition
Two matrices A, A0 such that there exist regular matrices P, Q satisfying
A0 = Q 1
·A·P
are called equivalent.
If two matrices are equivalent, then they have the same rank.
If two matrices represent the same linear mapping but in di↵erent
bases, then they are equivalent.
Conversely, two equivalent matrices represent the same linear
mapping, in di↵erent bases.

f i Un Vln
Af MBc.BZ Af MBnBe
Bs 132
MBap
Film Vm
Af MbcBe f
R Be
f Ralph Muz R
Egle µ f
Bce TM Be KEI EuE3iE4
flat bt7 9 ba
f t On
f Eat E3 Q 111107Be
t
A0
coordinates of Cdf w r Be
b I
1101111 Bc
fD
I
IG E Est EU
am
bio
A MIA
f t
iii it S0000
c I
f Htt ft lt
It
l 1,21pm
Mpaa to A
fH H III ci i i
osp
Homework is to check that
ii
t
MBe Ba
H p
Q
Definition
Two square matrices A, A0 such that there exists a regular matrix P
satisfying
A0 = P 1 · A · P
are said to be similar.
The notion comes from that of equivalent matrices, when Q = P.

If one considers a linear mapping f : V ! V (original and final
spaces coincide!!), and uses the same basis B in both cases, then
when B is changed to B 0 , the associated matrices A, A0 are similar,
and P is the change of basis matrix.
Conversely, two similar matrices represent a the same linear mapping
in di↵erent bases (related by P).

f RCH Pelt Bettie
Catb b alt
f Catbt
f HeP let lit Be FH t t C 1 1 Be

ai O
b
m A
B 1 It 2ft Basis
II HO YES
f l 2H i 3T Hi DB
flat H 3 t 11,17ps
MB pt LflJ
Homew
orkr To check that r
a
Ii'tT ati p
p I p
Kernel and Image
Definition
Let f : V ! V 0 be a linear mapping. We define:
(i) The nullspace or kernel of f , Ker(f ), is the set of all the vectors
of V that transform themselves into the vector ~0 2 V 0 .
(ii) The image of f , Im(V ), is the set of all the vectors of V 0 that are
the image of some vector of V .

Kernel and Image
Proposition
Ker(f ) is a vector subspace of V .
Proposition
Im(f ) is a vector subspace of V 0 , and its dimension is rank(A), where A
is the matrix associated with f (regardless of the bases used). In fact
Im(f ) = L({f (u1 ), . . . , f (un )}), where B = {u1 , . . . , un } is a basis of V .
Proof. Khan Academy (click)

Kernel and Image
Theorem
Let f : V ! V 0 be a linear mapping. The following statements are true:
(1) f is injective if and only if Ker (f ) = {~0}.
(2) f is surjective if and only if Im(f ) = V 0 .
Examples of how to compute Ker(f ) and Im(f ): Khan Academy (click)

Kernel and Image
Theorem
If f : V ! V 0 is an isomorphism and {~ u1 , . . . , u~n } is a basis of V , then
{f (~ un )} is a basis of V 0 .
u1 ), . . . , f (~

Endomorphisms
Definition
A linear mapping f : Vn ! Vn of a vector space onto itself is called
endomorphism.
The matrix associated with an endomorphism is a square matrix.

Conversely, any square matrix corresponds to some endomorphism.

Endomorphisms
Classification of endomorphisms: let f : Vn ! Vn be an

endomorphism, and let A be its associated matrix.
If |A| =
6 0, then f is bijective. In that case, we say that f is an
automorphism).
If |A| = 0, then f is neither injective nor surjective.

Endomorphisms
Proposition
Let f : Vn ! Vn be an automorphism, and let Af be the matrix of f .
Then f has an inverse, f 1 : Vn ! Vn , which is also linear, and its
associated matrix is A 1 .

Be Mani BA AB B Ebd
HI A VA
btc 0 dine VA 2
VAI atb d o
c i i
colours
it
profit I
Bashkimi
1 0
µ I
If I
ii vii
HI ri e
zI
rank hairs ei
ut Tee fi
are
III
T
is
Be us oil LUR Bc h e Bc
Mps
Bc It II CoidBa
Notice that I Air EVA and alee

A it EVA
Moreover became
ft L A
det HII A
j IL1 5 31 3
Homeworki Check that

AZ 3 At 31 0
I
3 II 3 A AZ AA II A f AA
At tI
Homework
as in
A L Yog the same
pop
revious example
the
14 102019
Problem 1. Given the following spaces of R4 :
Homos linear system
V1 = {(x, y, z, t) 2 R : x
4
y + z = 0, x z t = 0},
and
V2 = Span {(1, 1, 1, 0), (1, 0, 1, 1), (0, 0, 1, 1)} .
(a) Explain why they are vector spaces?
(b) Calculate the dimension and a basis of V1 , and V2 .
(c) Extend such bases to a bases of R4 , and call them 1 and 2 respectively.
(d) Obtain the matrix of change of bases from 2 to 1.
a VERY fixI c Vi
V C1124 Eva HeVs i
Iue Y
G
O
t H H c
yeR4
Vz satisfies a equation I
a µ
duh 4
2
2
Vi I III III li.eind.gr
X 9 z µ
9
Vi Spae 4,440 l co 1 1,1737 µ
p
D I 1 0
I
y 1 9
Mr µ
4
1 11,0 X
y 12 0
1,0 1 1 X z t o
t so
Cao lie z
free t 2
2 2
I H na a a
X 2 9
a
0
1,1 Z t 0
It 0 1 no
yt
Egs of Vz
i H t r
13 1
41141107,10511 411,610,407,610,9173
Vr
p t7
camp.pe ripping.pe II
Problem 2. Given the linear mapping f : R3 ! R3 defined as:
f (x, y, z) = (x y, y z, x z).
fled f 11,01071491
(a) Obtain the matrix associated to f .
(b) Is f injective? Obtain its kernel. fled fCon 07C1h07

(c) Obtain a basis of Im(f ). fled floral Co 4 1
(d) Given the bases
1 = {(1, 1, 0), (0, 1, 1), (1, 1, 1)} and 2 = {( 1, 0, 1), ( 1, 1, 0), (0, 1, 0)},
obtain the matrix of f with respect to the bases from 1 to 2.
l l 0
A
Ijn coma
t.ua
if KerCf
31019073 4 4,27 X y of
b f is injective Y 2 0
X 2 0
It
1 in o rrmamfhf.IE dinterCII
NI Kor f Spae UH
colours of A
c Im f is spanned by
f tho
Intf hat
µ I
span
Mlf
L L 5 L
Id M c
ftp.p.lt Mp
Linear Mappings
1. Let f : R2 ! R3 the linear mapping defined by
f (x, y) = (x + y, 2x y, 2x + 2y).
(a) Obtain the coordinate matrix of f with respect to the canonical bases.
(b) Obtain a basis, the dimension and a set of equations of the kernel of f and of the
image space of f .
(c) Obtain the coordinate matrix of f with respect to the basis B1 = {(1, 1), ( 1, 2)} of
R2 and of the basis B2 = {( 1, 1, 0), (1, 0, 1), (0, 0, 1)} of R3 .
a
flee f Io 112,2 i MBcBeH Y2
fled float It t.DE I eeBEoYheeIco
e
Mcf
X 4 C R2 f X y Co
oo
b Ker f
Xty 0
Equalise solution kerffimff.cof.o

z f o
and
1mCf Space
I ft l.ie iiEdimCImHY 2
AKO
Equations ZX Z o
cols
rank At solutions
c t
Mip.MCHMB.pe l i equations
MBiipzlfl MBCBZMLDMpspse
iii i'I little 1

HII't
2. Let f : R3 ! R3 be the linear mapping defined by
f (x, y, z) = (x + 2y z, y + z, x + y 2z).
(a) Obtain the matrix of f with respect to the basis B = {(0, 2, 1), (1, 0, 1), ( 1, 0, 0)}.
(b) Use this matrix to compute f ( 2, 2, 2).
IT 3
al
fFUe f 0,2 D 3,3 o Zz Z FB
I Ii.iooi i f IiIIse
i.on
E o
fl 32,4 11,1 3
B MpBtf Ig Ig 2Th GUI Gil's
bydef
0,4 47 10,4 4
2
3. Let f : R3 ! R4 be the linear mapping defined by:
f (1, 0, 1) = (1, 1, 1, 0)
f ( 1, 2, 0) = (1, 3, 0, 1)
f (0, 1, 1) = ( 1, 0, 1, 0).
Obtain the matrix associated to f with respect to the canonical bases and Ker(f ).
You need fled fled and fled Ji Choi

FEC1.2.07
a Are Kira Ble i B co h D
III If printer's
a
II
b You need MpsBc
Remember
MB.BEMido B
MBBc Miz B Gauss

Tordoff
process
i i ii n
p 1
I
Means
MizpahE MaffMpap
I I
3
1123 1124
f
ma Ei Ini
tanto Eat
Ker f x y 27 c
1123
f X y 27 6 o o o
o
Egs Xty
X 2g 122
0
II S
Ex KerCf Icao oB
No extensionof
Is f surjective why get an
cols A to get a
basis of R4
to i
4. Let f : M2⇥2 (R) ! R3 be the linear mapping defined by:
✓✓ ◆◆
I Ima
x y
f = (x + y, z, t).
z t
i
We want: i
(a) Prove that the application f is linear.
(b) Obtain a basis, the dimension and equations of the Ker(f ) and the image of f .
Weneedtocheck
take two elements
of Me CR felR
a
f AtB f A tf B
AIB
fCan afCA
d ffaIa4bdttaY lata'tbtb ctd Cdtd'D
ftp
g yflAtB atb d debt d d f A FCB
c
Laatab Ac dd A Cathg d
f GA f ta f
tfCA
I Karlf EE FLEEK to qoH fxty o z o t Eg of
spin to 073 theft
I'd I drunker L
O O O l
T
free din Ion di fl diulle.IE 4
I se
because He elemett
Golems of Mtf
Imff7
4
dimE 3 fatt 2 40 1
f
5. Let f : P (R)
2 ! R4 be the linear mapping defined by:
I 11
f a0 + a1 t + a2 t2 = (a0 + a1 + a2 , a0 2a1 , a2 , a0 ),
where E is the vector space of the polynomials of degree less or equal to 2 with real
coefficients. Find the coordinate matrix of f with respect to the bases
can
00
B1 = {1, t, t2 },
and
B2 = {(1, 0, 0, 1), (0, 1, 0, 0), (0, 0, 1, 1), (1, 1, 1, 0)}.
L d
FAY
Cbl Use such matrix to obtain f (1 + t).
My
O 111
A 20
in
idiotism
fetus 2
ii l l i
I
Ttt Clamps flittlethilots MB.BA hoI ftEzz
po4fomial Its coordinates Be
5
w.r.t.pe 124,010,1 0,4907 36,0141 314440
2 1
Lesson 4: Diagonalization
September 15, 2019
Eigenvalues, eigenvectors, eigenspaces
The rough idea: From the preceding lesson, we know that a matrix A
represents a linear mapping in a certain basis. On the other hand, the
matrix associated with a linear mapping changes when we change the
basis. So, maybe there exists some basis where the matrix is “specially
nice”...
Definition
Let f : V ! V be an endomorphism. We say that 2 R (or C) is an
eigenvalue of f if there exists ~v 2 V , ~v 6= ~0, such that f (~v ) = ~v .
Furthermore, in that case we say that ~v is an eigenvector associated
with .
Examples
7
How do we compute the eigenvalues:
DetIA AI
Properties:
(1) p( ) = |A
j
I | is called the characteristic polynomial of the
matrix A. Its degree is dim(V ).
dimA
(2) It is usual to refer to “the eigenvalues of the matrix” (values such
that A · ~v = ~v ), instead of the linear mapping.
How do we compute the eigenvalues: 2
Properties:
(3) If i is an eigenvalue, then it is a root of p( ), and therefore
ni
p( ) = ( i) ···
TO
The number ni is called the algebraic multiplicity of i.
(4) When we consider vector spaces over R, the eigenvalues can be
either real or complex.
Z
How do we compute the eigenvalues:
Properties:
(5) If is an eigenvalue of A and A · ~v = ~v , we say that ~v is an
eigenvector associated with .
Proposition
The set of all the eigenvectors associated with a same eigenvalue of a
matrix A, is a vector subspace.
Proof.
f i V V linear If dinka lvector SpaeeV
w vs
Ingeneral fr Vi A HI r o it is a space and
w wee ta i witwz e aw c a eIR
Definition
For each eigenvalue i , the set of eigenvectors associated with it is called
the eigenspace of i . We represent it by L i . From the above result, it
is a vector subspace, and its dimension is called the geometric
multiplicity of i .
dimLai
smg Ai
Observations:
(1) L i
is the solution of (A iI) · ~v = ~0.
(2) dim(L i ) = n rank(A i I ), where n is the order of A.
(3) Denoting the algebraic multiplicity of i by ni , it holds that
M 7 1  dim(L i )  ni
in
Example
Malai
MgGil
f PzCIRI IBAR
t e CE
flat btecE at b t Cbt2c f tfLH
e CH
a we
consider a basis
Pettit13
btk atb
a flatbtect fCc b a pa Cc pa
the matrixof fi
3 Weneed del A
det I
f ee fChocobo 1 2,07Pa o
mattoid
Ii feet 1
I Charad polynemili
delta
det att A 43 312 Hsp
Path
P f ripkmaG D 3
det CHIAl
k t Bakr
7G zu
WI A
fE og ithasrank
2
jmg.dz
MaCa M 1 A table
tMgla is NST diagonal
Theorem
Let f : V ! V be a linear mapping with p di↵erent eigenvalues
1 , . . . , p . Then the eigenvectors ~
v1 , . . . , ~vp associated with them are
linearly independent.
Diagonalization of a square matrix
Introduction: every square matrix A of order n is the matrix assoc. with

some endomorphism f : Vn ! Vn in some basis, i.e. A = M(f ; B, B).
Furthermore, if we change the basis B ! B 0 , then the matrix changes
according to A0 = P 1 · A · P. The question is: given A, is there any
basis where the expression of the corresponding endomorphism is
diagonal? In other words, given A, is it similar to any diagonal matrix?
Definition
Let A be a square matrix, and let f be the endomorphism that it
represents. We say that A (or f ) is diagonalizable if there exists some
basis such that the matrix associated with f in that basis is diagonal
(equivalently, if it is similar to some diagonal matrix).
Theorem
An endomorphism f : Vn ! Vn is diagonalizable if and only if there exists
a basis of Vn consisting of eigenvectors.
Proof.
ifand only if iff a
Theorem
Let V be a vector space over R of dimension n, and let f : Vn ! Vn be
an endomorphism. Then f is diagonalizable (over the reals) if and only if
the following two conditions hold:
(i) The total number of real eigenvalues, counting multiplicities, is n.
(ii) The geometric multiplicity of each eigenvalue equals its algebraic
multiplicity.
Jordan Matrix
In that case, there exists a matrix called Jordan matrix of A, which is
“block-diagonal”: 0 1
J1 0 ··· 0
B0 J2 ··· 0C
B C
J=B. .. .. .. C
@ .. . . .A
0 0 ··· Jr
where
• Each Ji (Jordan block) corresponds to a di↵erent eigenvalue i, and it

is a square matrix of order n i -by-n i of the form
0 1
i ? ··· 0 0
O
B0 ··· 0 0C
B i C
B .. .. . . .. .. C
Ji = B . . . . .C
B C
@0 0 ··· i ?A
0 0 ··· 0 i
and each ? is either a 1, or a 0.

Jordan Matrix
• The number and positions of the 1?s in each Jordan block must be
computed. We just mention the two “easy” rules:
1 If dim(L i ) = n i , then the Jordan block is diagonal (no 1?s);
2 if dim(L i ) = 1 6= n i , then the block has 1?s above all the
elements in the main diagonal;
so, in that case it looks:
0 1
i 1 ··· 0 0
B0 ··· 0 0C
B i C
B .. C
Ji = B ... ..
.
..
.
..
. .C
B C
@0 0 ··· i 1A
0 0 ··· 0 i
Jordan Matrix
It also happens that where P fulfills: I regular

1
A = PJP
where P fulfills: b
D
The columns of P include independent eigenvectors (but we need
more!!).
The remaining, unknown, columns must be computed:
(i) by using that AP = PJ (not very efficient...);
(ii) using a more sophisticated/efficient method that we will skip here
(but it exists!)
let A stuITIFEis diagonalaide
L 3g
Obtain matrices E and D so that A PDP
PaHgfYI
detHI
A
Igf.ca alto f cancan
hast
simple A diagonalizable
tf
Ei J
E ol40io7
diI A f8oz9IJ z7yIIIofx
dz
is c z
wIa.f
iEat aor
fitriB I
I E
Exempt
Let a c IR Study for which values of a the matrix
30
A is diagonalizable
that A PDP
matrices E and D so
Obtain
alI at
a a
Hal O t
EE o.atca u n
f I 1 2 da I
if
d d3
a a
Simple A dagou.at
42 2
a
malt 2
if a p 4
12
1
2
double
simple Matt
21 1 I A is not
diagardizable
h I hide math A is NOT

if an
12 2 double Mala 272 diagonalitable
rank is 2
d I A L
Ar 17 3 2 1
Mg
rank is 2
12 detI A
2
L fq
Mg 22 2 3 2 1
a 1.2 HI A ooo rT Uio
a
d HI A
f a
Felsite
rI C3o
d WI A
Ag
Iz
a
I L e too
Ii E Ii
E
A PDP A PDP157 67 D P
An
P t EDwp
ME IN
a nE a
E
i
6. Let E be the space of polynomials of degree at most 2 with real coefficients and let F
be the 2-by-2 squared matrices with real entries. Let us consider the linear mapping
f : E ! F defined as follows:
✓ ◆
p(0) p(1)
Heat
2
f (a + bt + ct ) =
p(1) p(0)
pie
(a) Obtain the coordinate matrix of f with respect to the basis
B1 = {1, 1 + t, 1 + t + t2 }
kid
Be IEtiEqE3 Ey
of E and the canonical basis of F . fput.su
(b) By using the matrix obtained in (a), compute f (p) where p = 3 + 2t + t2 .
(c) By using the matrix obtained in (a), compute a basis of the subpsace Ker(f ).
ftile flip I f e FE tEzHE3HEY 1,11 1 Bc

MBBc f
a TTU
set b o
c o
fl ffya I LY EItU fh2HE
set b I
c to
ftp.ffht ttI z7
EieBEz
a I
3EstEI3y3DBc
set 5 1
c I
barisBy
w.at
pH
msn.tt kid need to
express
f
uiMlB lb7fCp7
ptH 3t2tt eu ai
x.stpCkHt8Htt
system
solve the
3tzt lt ktp N lptbt Btr
IT 6 rtg.atper 8 1
ftp.MB.BIHHI fqk.bg fhI 1383
35
of matrices
a collection
polynomial became e c Blk
c Ker is of
era
E
A 0 O L
2
r n l
KerCfl Span HEI free t
7. Let E be the 2-by-2 squared matrices with real entries and let F be the space of polynomi-
als of degree at most 2 with real coefficients. We consider the linear mapping f : E ! F
defined as follows: ✓✓ ◆◆
a b
f
c d
= a + (b + c)t + dt2 .
c f 2t3t
(a) Obtain the coordinate matrix of f with respect to the basis
fieii.IEoff
⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
B1 =
1 0
0 0
,
1 1
0 0 co
,
1 1
1 0
,
1 1
1 1
iii
of E and the canonical basis of F .
at ios ice
✓ ◆
2 2
(b) By using the matrix obtained in (a), compute f (C), being C =
1 0
.
Etui
Cal f UT f f8 hot107 1 05 6,9 1 Be
T TB Coil 1 Be
a I f UI It ttoE let
set
get fat tl f i zt ioE zr iB C92
d o
Cd
at 57
c I
d0
2
filly f ff 1 21 1 1 2 03 11,21 Be
at b I
a7 TV
c I
µ
pe fg9II
Anemia se n cam a
fIIII3l
b C Hofstetter491,101,3
7
f c 8,9 4 g
E Bc 3 2
8. Let E be the space of 2-by-2 matrices with real entries. We consider the linear mapping
f : E ! E given by:
f (A) = A + AT ,
where AT is the matrix transposed of A. AIBT
(a) Prove that f is a linear mapping.
HA iB CA B CAIBYICA AH CB.pt
(b) Obtain the coordinate matrix of f with respect to the basis
⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ fCAlefCB7
B=
1 0
0 0
,
1 1
0 0
, O
1 1
1 0
,
1 1
0 1
Dd
FAA GAltGAT
of E. e
Us og ✓
2 3
◆
dAtdAT
(c) By using the matrix obtained in (b), compute f (C), being C = .
O
2 1
d AcAT
p
1,0 2,1 B tfCA
b
ftp.pdfl MB.BEMCDMB.Bc
e i l b
let me also reunber fluit UT UT f.IE 2UTtou7tou3eouy
2,0101dB
fcuj.u ieu I ffffIH.tk that.de
I g i u ti i
2 I I 0
msn.at
t H
8
d Gi L UT aatzt.tle t1iQ2iDg BfC h0i2il7B MB.B f Ig
I
t 13311341 figs if f 1 4 4921,3
C Th 2 505 54 14552
af pen Hatti L feat

7
Mala07
h o simple
MaG 2 3
12 2 triple
glad I A L rank ist

mga 22 4 1 3 d.tt
a
l Looker COLI A o o 2t 03
zx
ytz
Spanko l 1107
20
12 Ker ZI A ytZ of
010,0
Span Choco d 0,111,0
toilet Eia're
A PD
9. Let E be the space of polynomials of degree at most 2 with real coefficients. We consider
the linear mapping f : E ! E given by:
f (p(t)) = p0 (t),
ffatbt ct22
(i.e. the image of a polynomial is its derivative). btzc u.ie
ie ca7fluTl
(a) Obtain the coordinate matrix of f with respect to the basis B = {t, 1 + t, t2 } of E.
(b) By using the matrix obtained in (a), compute f (1 + 2t + 2t2 ).
flH I tUIIi f.lusI flt7 2t 2YI

ao UfoB get a 0 2,40 B
ft bei 5
Ceo c I
fluids filth L uIuT tailors

at Ei
Mps.rs
fI to
b fllt2t 2EI fthhDB MBBCflfff.LY 2u7t2kz

p 2t 2HtL
2
2 zt2
t xu.HU ru3 4 1ple Lt t rF p ldtB7t N
Z
f I 2 1 2
f 1,2
tf
Study is diagonalizable
faff 1
ItII f tYIt a.d2 d3f
9
pttt detlNI.AE d Fpe maCd

o 3
Lo KerloE At Xty Zzz
to JF IT8o e
X
y 0
Equating
dim 6 3 2 1 Mgl'd07 1
Mala07 1 Mgb07 A is NOT dragonalizable
NOT FOR THE EXAM
WARNING
spaett 1.073 totter out AT FILI
spanking 113
e
FALI
3 1123
Karlo A
co oil
span411 407,111,07 Burt
Ear Is
riv
at Ip
i
i be
10. Let E be the space of polynomials of degree at most 2 with real coefficients and let F be
the space of polynomials of degree at most 3 with real coefficients. We consider the linear
mapping f : E ! F defined as follows
Z t
BE flatbt ct7 attbtz.cc
It
f (p(t)) = p(s) ds
0
(i.e. the image of a polynomial is its definite integral between 0 and t: a primitive of the
polynomial so that at 0 is equal to 0.)
(a) Obtain the coordinate matrix of f with respect to the basis B = {1, 1 + t, 1 + t2 } of
E and the canonical basis of F .
(b) By using the matrix obtained in (a), compute f (3 + t + t2 ). es
Cal flatlet t r5 Co o 1,07
Be fatd Shtetl
la
t tz.fr tB
a I
St boo set b I 0,12cL07
Ceo Cto
ftp.flt E tttz3 z1JitFPs 1310,40Be

set a I
boo
g
Mps
L Ff
b
flzyfet7 flhhNB MB.B.CH
I fz jEEtfEt3t
w need to writeit w r t B
3 it it2 dhi pfz tuz x.lt putt 18 little ftp.IJ ptttt
dtBtt 3
a
I
10
411
Lesson 5: Linear Di↵erential Equations
September 29, 2019

Introduction
A general problem in Applied Mathematics is the following:

We have two variables, y and t, which are related, i.e y = y (t), and
we want to find the function y (t).
Nature, in general, does not make it easy to find, directly, y = y (t).
Instead, it usually provides a certain relationship between the
derivatives of y (t) with respect to t, the function y (t), and t that
we have to exploit to find y = y (t).
Examples: planetary movement, elasticity, radioactive decay,
population dynamics, circuits, . . .

Introduction
uh until
L
iit
R
t
L
Eun
C
n
E Ult p
L
d 2 q(t)
dt 2
+R
11111 we solveHe
dq(t)
dt
1
+ q(t) = E .
C
in
weassure
it Lg 4
Walt
E Rualt U It
E is constant
system
Introduction
was
EI ut
I
1H i1
8
> i10 (t) = i2 (t) 2i3 (t)
p in d II If I
>
>
1H 1H <
i20 (t) = 4i2 (t) + i3 (t)
>
>
>
: i 0 (t) = 2i (t) 6i (t)
2 3
ft lo
3
t 22
i2 3⌦ i3
1⌦ 2⌦ T

Introduction
So, the first problem is: given F (t, y , y 0 , y 00 , . . . , y (n) ) = 0,
can we find y = y (t)?
Difficult, and not solvable in general.

Here, we focus on some treatable cases, whose study is related to
the topics addressed in previous lessons.

Di↵erential equations
An ordinary di↵erential equation (o.d.e) is an equality involving an
unknown function y , and its derivatives with respect to a certain
independent variable t:
F (t, y , y 0 , y 00 , . . . , y (n) ) = 0 = 0
Examples:
y 0 (t) = y (t) YIH et
✓ ◆
0 y (t)
y (t) 2y (t) y (t) =0
10
✓ ◆
y (t)
y 0 (t) = 2y (t) t 2
10
y 00 (t) sin(t y 0 (t)) + y (t) = 0
..
.
The order of an o.d.e. is the order of the highest derivative appearing in
the equation.
How do we compute the eigenvalues?

• An explicit solution of F (t, y , y 0 , ..., y (n) ) = 0 is a function y = (t)

fulfilling the equation, i.e. satisfying that
0 (n)
F (t, (t), (t), ..., (t)) = 0
Examples:
yIH Ket K
Solutions of y 0 (t) = y (t)? constant
y.lt Kcosit y Kzsink

Solutions of y 00 (t) = y (t)?
Check that y (t) = e t is a solution of y 00 (t) 2y 0 (t) + y (t) = 0.
I
check as wellthat
solution
yettet is a

An o.d.e. always has infinitely many solutions.

A solution of an o.d.e. is called a particular solution.
The general solution of an o.d.e. of order n is an expression,
depending on n constants, which provides infinitely many solutions:
(t, y , C1 , . . . , Cn )
(examples relative to the o.d.e’s in the previous slide).

Initial Value Problem (I.V.P.) It is an o.d.e. together with some

initial conditions referred to a same value of t.
Boundary Value Problem (B.V.P.) It is an o.d.e. together with
some conditions referred to di↵erent values of t.
Both I.V.P. and B.V.P may or may not have solution!! Furthermore,
if it has a solution, it may or may not be unique. In general, we will
look for the solutions by imposing the initial conditions on the
general solution.
yea a g pau wa µ aye

etc

Here, we will restrict ourselves to the easiest case, namely that of

linear di↵erential equations:
an (t)y (n) (t) + an 1 (t)y

(n 1)
(t) + · · · + a0 (t)y (t) = b(t)
In the above case, we say that the equation is homogeneous if b(t)

is identically 0. In other case, we say that it is non-homogeneous.
If ai (t) = ai for i = 1, . . . , n, we say that it is linear with constant
coefficients.

• Linear di↵erential equations have the following nice property:
Theorem
Let an (t)y (n) (t) + an 1 (t)y (n 1) (t) + · · · + a0 (t)y (t) = b(t), and let
J ⇢ R be an interval where all the functions ai (t)’s and b(t) are
continuous. Then, for any t0 2 J, and any choosing
(n 1)
y0 , y00 , . . . , y0 2 R,
the Initial Value Problem

8
>
> an (t)y (n) (t) + an 1 (t)y
(n 1)
(t) + · · · + a0 (t)y (t) = b(t)
>
>
>
>
< y (t0 ) = y0
y 0 (t0 ) = y00
>
> ...
>
>
>
>
: (n 1) (n 1)
y (t0 ) = y0
has a solution, and it is unique.

generalresult

First Order Linear Di↵erential Equations
First order linear equation: The general form is
y 0 (t) + a(t)y (t) = b(t)
We consider first the homogeneous case:

at k
y 0 (t) + a(t)y (t) = 0
yIH e l kconst
rameter
Theorem
The set of solutions of an homogeneous first order linear equation has a
structure of vector space of dimension 1.

First Order Linear Di↵erential Equations
Non-homogeneous case
y 0 (t) + a(t)y (t) = b(t)
Theorem
Denote by ynh (t) the general solution of the above non-homogeneous,
first order linear equation, denote by yp (t) a particular solution of the
above non-homogeneous equation, and denote by yh (t) denotes the
general solution of the associated homogeneous first order linear
equation. Then it holds that
ynh (t) = yh (t) + yp (t)
Exanpkixty.se NH Parti XP t 410 0

D µn µy p x
x y o CH select Yuu d Yp
eye
Yu d Lesson 5: Linear Di↵erential Equations
Second Order Linear Di↵erential Equations
The general form of these equations is:
y 00 (t) + p(t)y 0 (t) + q(t)y (t) = b(t)
As before, if b(t) = 0 we say that the equation is homogeneous.

Theorem
The set of solutions of an homogeneous second order linear equation has
a structure of vector space of dimension 2. So, the general solution of
such an o.d.e. is
yh (t) = C1 y1 (t) + C2 y2 (t),
where y1 (t), y2 (t) are linearly independent solutions.
t
whatdoesthatmean

So, whenever we find two independent solutions y1 (t), y2 (t) of the o.d.e.,
we get the general solution.
How do we check if the functions
y1 (t), y2 (t), ..., yn (t)
are linearly independent?

Higher order Linear Di↵erential Equations
Definition
The Wronskian W (y1 , y2 , . . . , yn ) of y1 (t), y2 (t), . . . , yn (t) is the
determinant
y1 y2 ··· yn
y10 y20 ··· yn0
W (y1 , y2 , . . . , yn ) = .. .. .. ..
. . . .
I
(n 1) (n 1) (n 1)
y1 y2 ··· yn
defterminant
Remark: Notice that W (y1 , y2 , . . . , yn ), in general, is a function on the

variable t.

Higher order Linear Di↵erential Equations
Proposition
If y1 (t), . . . , yn (t) are linearly dependent, then W (y1 , y2 , . . . , yn ) = 0
(i.e. it is identically 0).
Tense
Consequence: if W (y1 , y2 , . . . , yn ) is not identically 0 (i.e. it might be 0

for some t’s, but not for all t’s!!), then y1 (t), . . . , yn (t) are linearly
independent.

Case of constant coefficients: we start with the homogeneous
y 00 (t) + a1 y 0 (t) + a0 y (t) = 0
Ka t ao polynomial of2nddegree
J
Characteristic equation and discussion
.
youneed replace y
y A
y 5 andsolvethe
equation
y1,5
I a d do 0
to b a.tt RLI circuit
a
2 Example
h a F4ao c a
then the solutions are

If he be
constants
ks.kz
t ed.tt ttdzeHt2dt d ecaitda

wlyiiyzl fy.ee eine ytzeaztf
eHHHt It th da to
This is o
only if 4 1 da o but hehe
which means he 6
If di k d tea
Y IH ed't k kL Kz constants
yzltTe.tk
why ms
fIIteIetht e tch o
t
if I th b Ed h dtj dz jµJ
I9 unawares e
i
Non-homogeneous case:
y 00 (t) + p(t)y 0 (t) + q(t)y (t) = b(t)
Theorem
Denote by ynh (t) the general solution of the above non-homogeneous,
second order linear equation, denote by yp (t) a particular solution of the
above non-homogeneous equation, and denote by yh (t) the general
solution of the associated homogeneous second order linear equation.
Then, it holds
ynh (t) = yh (t) + yp (t)

1
Diagonalization
1. Study if the following matrices can be diagonalized over R and, if so, obtain matrices D
(Diagonal) and P (regular) so that A = P DP 1 .
0 1
0 1 0 1 1 0 1 1
1 1 1 1 0 0 B 2 1 2 4C
(a) A = @0 1 0A (b) A = @1 1 2A (c) A = B @ 1
C.
1 1 1A
0 1 2 0 0 1
1 1 1 3
cat phd det HI A A 1 d Htt 2 1 1 Mala D

6 2 Ma 2 2 1
4 Kor AI A
NI A
I f dim 4 mgca.si 3 2
A is not diagonalizable
raid
b play d DA DA D maH
A is not
L KerCUI A diagaalizable
f g 4 MgCd 1 3
dim 1 2
NI A
rank 1
ant it i L
Ha
ith II ah EH 27 mm 95.7
A 1
2 2 1
µMgtd rau.kz Ma
Ii 41 AI A Lf Ais DIAGONAL2ARE
Span Gioin1,1 Ct His
columns
1 0,07 free
L
LEE
is
o
at II I
Lo span U o t o
to
ix y
yet
o
o
o
a
t.tt p
04
free
spanks to.ms
74
af IIH 7iII4nfoo oEIo
002
1 9 81
fririrtriH
2. Let f : R3 ! R3 be a linear mapping which has the following matritial representation
with respect to the canonical basis:
0 1
1 3 0
A = @0 a 0 A .
2 1 1
(a) Find the values of a so that A is diagonalizable over R.

1
(b) For a = 3 obtain the matrices D (diagonal) and P (regular) such that A = P DP .
Al
all It
plat det CHI
II f
can a
The Spectrum is TCA L I t a

d 1 ATL d a
Discussion i
I then A has single eigenvalues
if Fs nora a t
is diagonalizable
therefore A
ma 117 2 In such a case we need to compute
if a L then
ma 1 17 1 MgCD
A Mach
MgIN dim Le L Ker AI
dim
AI NI A
f E rank z
so A is not diagonalizable for a L
2 In such a case we need to compute

If a l then Mac 17
mall 1 myth
Kerl NI A malt
MgH7 dim le L
diml.ie
so A is
a I
not
III A
diasonalizable
I 2
I
a
f
1 either
rank z
A 3
D
f is e't.feteoe.ge O O
E
III II a I's
TEI II o.o.se
um e
PeltDDp i
un
uu.is
II Ile9eIi'eiiifp uiD
f
it ulk cas.at
i i It
it N
y O 0 A 1
u.nl 3
IE 3l
e
eii IfiIeiet IE
3Ettt
3 2
3e
3t3t
1z e
ltt 313 LI
d t.ze tze
solution A
f youwant to check
your owf
z
It
µ Ub 24th Uz
3. Let us consider the matrix 0 1
b 0 2b
A= @ 1 1 2 A
b 0 2b
that depends on the real parameter b.
a) Determine the values of b so that A is diagonalizable over R.

1
b) For b = 0 find the matrices D (diagonal) and P (regular) so that A = P DP .
c) Compute, As application of (b), the power An (for the b = 0 case) for any positive
integer n.
t
a pad detCHI A Itb f.bz CaD bbf7zb laD f 3rb
dH DG 3D of 011,333 Ep yA_
because
Discussion if 3b 1 0,1 then A is diagonalizable
A has 3 single eigenvalues
b then Maco 2
up if 3b 0 o
Machel
2 1
AIL
Lo Ker coli A dim6 3 1
A
Iggy
A If 3b I be43 there ma o 1
Mall 2 NOT
A is
Ker ttt A then dim Lee 3 2 1 diag
it a
I
solu t.A ryb.withb.tl
3
b BIO to Kercott A x y 27 0
Span 141,07 2,011

in ie
Lasker NI A L to
span40,4073
2 0 IT
e
NI A III
a
I I O d
Cc b i 1
µ
a
ear't D io lio
ItioIII
l t
l
i tsE.iIy
1
i H to i i
i aint
4. Let f : R3 ! R3 be the linear mapping defined by:
f (x, y, z) = (↵x ↵y + 3z, 2↵x + 2↵y + z, 3z).
where ↵ is real.
(a) Compute A the coordinate matrix with respect to the canonical basis.
(b) Determine the values of ↵ so that A is diagonalizable over R.
1
(c) For ↵ = 1 find the matrices D (diagonal) and P (regular) so that A = P DP .
(d) Compute, as application of the precious result, the power An (fot the ↵ = 1 case)
as a function on n. Simplify as mush as possible.
4
Linear Di↵erential Equations of Order n
General form:
an (t)yn (t) + an 1 (t)yn 1 (t) + · · · + a0 (t)y (t) = b(t)
Homogenous case:
an (t)yn (t) + an 1 (t)yn 1 (t) + · · · + a0 (t)y (t) = 0
Theorem
The set of solutions of an homogeneous linear equation of order n has a
structure of vector space of dimension n. So, the general solution of such
an o.d.e. is
yh (t) = C1 y1 (t) + C2 y2 (t) + · · · + Cn yn (t),
where y1 (t), y2 (t), · · · , yn (t) are linearly independent solutions.

Non-Homogenous case:
an (t)yn (t) + an 1 (t)yn 1 (t) + · · · + a0 (t)y (t) = b(t)
Theorem
If ynh (t) denotes the general solution of a non-homogeneous linear
equation of order n, yh (t) is the general solution of the associated
homogeneous equation and yp (t) is particular solution of the
non-homogeneous, then
ynh (t) = yh (t) + yp (t).

Constant coefficients, homogeneous case
an yn (t) + an 1 yn 1 (t) + · · · + a0 y (t) = 0
Characteristic equation:
an r n + an 1r
n 1
+ · · · + a0 = 0
As in the case n = 2, each root rk (real or complex), of multiplicity
mk , gives rise to mk independent solutions:
Every real root rk of multiplicity mk gives rise to
e rk t , te rk t , . . . t mk 1 rk t
e .
Every pair of complex roots ak ± jbk of multiplicity mk gives rise to
e ak t sin(bkt), te ak t sin(bk t), . . . , t mk 1 ak t

e sin(bk t)
and
e ak t cos(bkt), te ak t cos(bk t), . . . , t mk 1 ak t

e cos(bk t)

First Order Systems of Di↵erential Equations
A first order system of di↵erential equations is a collection of equations

8
> du1
>
> = f1 (t, u1 , u2 , · · · , un ),
>
> dt
>
>
>
> du2
< = f2 (t, u1 , u2 , · · · , un ),
dt
>
> ..
>
>
>
> .
>
> dun
>
: = fn (t, u1 , u2 , · · · , un ),
dt

If, in addition, we impose some initial conditions, meaning

8
>
> u1 (t0 ) = u10 ,
>
>
>
>
< u2 (t0 ) = u20 ,
>
> ..
>
>
>
> .
:
un (t0 ) = un0 ,
then we have an IVP.
EXAMPLES

What is now a “particular solution”?
8
>
> u1 (t)
>
>
>
>
< u2 (t)
>
> ..
>
>
>
> .
:
un (t)
What is a “general solution’”?
8
>
> u1 (t, C1 , . . . , Cn )
>
>
>
>
< u2 (t, C1 , . . . , Cn )
>
> ..
>
>
>
> .
:
un (t, C1 , . . . , Cn ).
As many constants, as equations.
EXAMPLES
Transforming Equations of High Order into First Order
Linear Systems
An o.d.e. of order m (not necessarily linear) of the following type can be

transformed into a first-order di↵erential system:
y (m) (t) = f (t, y , y 0 , . . . , y (m 1)

)
Important
in the following way:
9 8
u1 = y >
> >
> u10 = u2
>
> >
>
>
> >
>
u2 = y0 = < u20 = u3
)
.. >
> >
> ..
. >
> >
>
>
> >
> .
; : 0
um = y (m 1)
. um = f (t, u1 , u2 , . . . , um ).
EXAMPLE

Linear First Order Systems of Di↵erential Equations
Consider just two functions (i.e. n = 2). A first order system of

di↵erential equations with two unknowns is linear, if it can be written as
8
< x 0 = a1 1(t)x + a12 (t)y + b1 (t)
:
y0 = a2 1(t)x + a22 (t)y + b2 (t).
Denoting (x, y ) by u, (b1 (t), b2 (t)) by b(t), and

✓ ◆
a11 (t) a12 (t)
A(t) = ,
a21 (t) a22 (t)
we get
du(t)
= A(t)u + b(t).
dt

In general, we have
du(t)
= A(t)u + b(t),
dt
where
A(t) is an n-by-n square matrix whose elements are functions of t.
b(t) = (b1 (t), . . . , bn (t)) is a vector whose n coordinates are
functions of t.
u = (u1 , . . . , un ) where ui = ui (t) for i = 1, . . . , n.
If, in addition, we have initial conditions u1 (t0 ) = u10 , . . . , un (t0 ) = un0 ,
denoting (u10 , . . . , un0 ) by u(t0 ), then we have the following IVP:
8
< du(t) = A(t)u + b(t),
dt
:
u(t0 ) = u 0 .

Given
du(t)
= A(t)u + b(t),
dt
we say that
Is homogeneous if b(t) = 0, and non-homogeneous otherwise.
Has constant coefficients if A(t) = A, i.e. it is a constant matrix.
In the sequel we focus only on systems with constant coefficients.

Linear First Order Systems of Di↵erential Equations with
Constant Coefficients
Homogeneous case:
du(t)
= Au.
dt
If we think of
dy
= ay ) y (t) = Ce at .
dt
So, could we try something like u(t) = Ce At ?
Does this make sense?

series
Introduction: Taylor’s polynomial.
EXAMPLES
et it t I Ist
sint t
t
tf t

Introduction: Taylor’s polynomial.

The exponential matrix
Let A be a n-by-n matrix, the exponential matrix l of A, e A is another
matrix defined as follows:
A A2 An
eA = I + + + ··· + + ···
1! 2! n!
Observe that then
A2 t 2 An t n
A eAt
At
e At = I + + + ··· + + ···
1! 2! n!
Here n! represents the factorial number defined as n! = 1 ⇥ 2 ⇥ · · · ⇥ n.
2
feat Ay ftp.h AZz i A I Ant A E
If
A = Diag( 1, 2, . . . , n ),
then
e A = Diag(e 1 , e 2 , . . . , e n ).
Properties:
e 0 = I,
If A and B commute, then e A+B = e A e B .
I=entries
1
The matrix e A is regular and e A e . A
The matrix e At is di↵erentiable (all tis are di↵erentiable), and
d e At
= A e At .
dt

Theorem:
The general solution of
du(t)
= Au.
dt
is
u(t) = e At C .
How does the solution look like when n = 2?
EXAMPLE

Theorem:
The general solution of the Initial Values Problem
8
< du(t) = A(t)u,
dt
:
type i toro
u(t0 ) = u 0 .
type to 20
is
u(t) = e A(t t0 )
u(t0 ).

PROBLEM: how do we compute e At , in the A non diagonal case A?

If A is not diagonal, but it is diagonalizable, then we use
TT eat pept E
1
A = PDP
If A is not diagonalizable, it can also be done (eigenvalues help also

here), but further questions are necessary (Jordan forms).
skipthis
case

du(t)
dt
= A(t)u + b(t),
D
Again, in this case we have
u nh (t) = u h (t) + u p (t).
We know how to compute u h (t); in fact we have
u h (t) = e At C .
In order to find u p (t), we use the variation of parameters method:
we look for a particular solution u p (t) = e At C (t), where C (t) is an
unknown (vector) function.
By imposing that u p (t) is a solution of the equation, we obtain
Z
e
C (t) = e As b(s) ds.

Hence, the solution of the equation
du(t)
= A(t)u + b(t),
dt
is
u(t) = e At
C +e
dZ
At
e
db(s) ds.
As
If we add the initial condition u(t0 ) = u 0 , then
u(t) = e A(t t0 )
u0 + e At
Z t
e As
b(s) ds.
IfPellitory
t0
type Eton

2
PICHI f E E
A MA 14 12124
in It if it
13 1881,1 11 1,1117
a MBB f MBBeCf
i
Nd
b f C 1423
dkerCt
a flute Mf 1 to antipathies Spy
1 2,210713
fluid M e
LL to 1,0 2
B
IoIEIo b
Ma.at
fCc7 flIiIihMB MB.plflf
f o p
suis Huit6uttloties
L
c Ker
_spank HELLO 39dB span ffg fff
free 2 D tell
I
a team
f 022T p 22 24
b 2 0
A = @2 b 0 A
1 0 2
a) Study the values of b for whose the matrix A is diagonalizable over R

1
b) For the b = 1 case obtain matrices D (diagonal) and P (regular) so that A = P DP .
c) Obtain, as an application of the previous result, the power An (in the b = 1 case)
for any positive integer as a function on b. Simplify it as much as possible.
A
f f pal delCHIAl I Eobf lazy Ebt
h2 T 2bItb 4
in Jae zbt2 b 2
b1 0 b
d 1227
Dim If b 12 2 di b 2
nor b 2 1 2 3 diff eigenvalues so A is
t diagonalizable
b 14
if be 4 malt21 2
mat 6 1 go A is Not
dragonalizable
z Ker t2I A dim mg 2 L
ZI A fFgfD2 rank is 2
If bio Mac27 2
to A is not
Mate 1
dragonalizable
KertzlI A dine L
t d my
t g
a ZI A
f rank is 2
b
A
IIE ra 12 133 5
Lz Hrc 4 Kerl spanks LIED
zqqgtqty
spanlloiof.fi i
IIIA spank D
EEE o you can
multiply
Km HI A Span 441 a constant
T
ri
pego
D go E fog
71 bet
man PDH
III I III
i xi t
i ssi.iI
fi io
2nH fiTt3nH 4.2n tiYt3nt 6.2
a 1 1
A = @ 1 a 0A
0 0 1
that depends on the real parameter a.
a) Study the values of a for whose the matrix A is diagonalizable over R

1
b) For the a = 1 case obtain matrices D (diagonal) and P (regular) so that A = P DP .
c) Obtain, as an application of the previous result, the power An (in the a = 1 case)
cal Platt T.IE caafIiaIat72HhlI zahea2D

a1 0 a att
att 11 l11
Dassin if nor a
a
then A is Diagonalizable
Spectrum TA 1 atlatl
off
the male L Kor NI A dim4 Mg L
If a 0 2
It
Mac17 1 1
Mals
so A is NIT diagonalitabh
ma 1 ez
f to
000
If 9 2 then
Li Kor NI A dim4 L
Ma 37 1
1
ng It
Matt
so A is Not diagonalizable I 0
000
b a 1 A
f ra 1,0 23
Li Ker NI A Span 0,1 17 Lo KorCoE A Spank 1,07

I Va
t II
Eo
6
Kor 3 II Al Spank2 1 OD
i J
I L HII II
47 Homework
1 0 3
@
A= 0 2 0A
0 2 a
that depends on the real parameter a.
a) Study the values of a for whose the matrix A is diagonalizable over R

1
b) For the a = 3 case obtain matrices D (diagonal) and P (regular) so that A = P DP .
c) Obtain, as an application of the previous result, the power An (in the a = 3 case)
Homework
pin HI Al Ez 301
ca
I O 2 da
ca na Alta
of 21,3A
A has 3 diff single
i on
If a s nor a z gthee
eigenvalues hence A is diagonalizable
If a l then Mal Dez KaracIA dim 4 Mg Dz

Ma 2 p t
ratukkisz male
Q A is NET drag
dim Length
If a 2 then Ma127 2 Lz terCUI A
1 Ma
ma ut l
88 8
020
So A is NOI diag
7
1 0 0
@
A= 0 3 2A
2 0 b
a) Study the values of b for whose the matrix A is diagonalizable over R

1
b) For the b = 2 case obtain matrices D (diagonal) and P (regular) so that A = P DP .
c) Obtain, as an application of the previous result, the power An (in the b = 2 case)
Homework
to
O O
a INI Al a 179 37 la b the spectrum is
Jodtzz
2 O d I 3,54
TA
Discussion If b1 1 nor bt 3 Ha
A is diagonalizable
L Kor NI A dim 4 1 Mals

if b t then male 2
t
Mac37 1
EEE
so A is notdiagonalized
KarGII A dim 13 1 Ma f 3
if b 3 then male 1
t
mat37 2
So A is Notdiagonalizable
f If200If
8
Linear Di↵erential Equations
EASY
1. Solve the following initial values problem: I
(
Type 0 1
u 0
= Au 0 1 1
where A = @1 0 1 A .
u(0) = (1, 1, 1)T 1 1 0
A
f platini At
HII f B i i a t t 1332 2
i 42 2 2 2 5421 1
Z method 43 32 2
A O Z Ruffini's T
KIIotinthAt a swoosh
C TA 42 4 13
El t I
ma E z
ri iz
4 Ker GI t A Spark 07,111017
Jithas ranks Mg67 2 A is DIAGONAL'ZABLE

1,1
kwC2I A Hank Fff
43212
D
E Et
00
i
Is tf tf
1
settle eat Uco d
Solution PeDZPI
o
to i e
ii iii
S
iii i i l it
solution nut III
wait H II r
Lesson 6: Euclidean Spaces
October 27, 2019

Motivation: (1) GPS Technology
We want to find the coordinates (x, y , z) of the point where we are. Let
us address first the problem in the plane (we want to find (x, y ))

We call a satellite. The satellite “knows” where it is; also, by calling the
satellite, we know the distance from the satellite to us. So, we get a
circle, and we know that we are a point on the circle.

If we call another satellite, then we find two circles containing us. So,
two possible solutions. But one of them can be easily discarded.
Discarded

But in fact there are many more satellites, so we can call them, and get
information from them also. In fact, when we sort to the 3D case we
need at least three satellites to find our position as the intersection of
three spheres (this provides two solutions, again one is discarded).

The satellites are very far away. So in fact, in the vicinity of our position
the circles can be approached as lines. Equivalently, the spheres can be
approximated as planes.

In an “ideal” setup, all these lines (planes) intersect exactly at our

position. But in the real world, any measure carries a certain error. As a
consequence, these lines (planes) do NOT intersect at a point, although
certainly they are “close” to intersect exactly.

Since we need to intersect planes, we need to solve a linear system of

equations:
8
>
> a11 x + a12 y + a13 z = b1
>
>
>
>
< a21 x + a22 y + a23 z = b2
>
> .. ..
>
>
>
> . .
:
am1 x + am2 y + am3 z = bm
where m is bigger than 3. But, due to errors in the measures, this

system (see the previous slide) is not consistent. Still, it is very “close”
to being consistent... So, does it make sense to look for something
that we can call a “solution” of this unconsistent system?

Motivation: (2) Fourier Series
Given a signal f (t), we want to write it as:
f (t) = a0 + a1 sin(t) + a2 sin(2t) + · · · + b1 cos(t) + b2 cos(2t) + · · · .
In other words
1
X
f (t) = a0 + [an sin(nt) + bn cos(nt)] .
n=1
Rough idea: the signal f (t) is decomposed into simpler signals, with
increasing frequency. Often, the most relevant information on f (t) is
carried by the terms of small frequency.
How do we do this, i.e. how do we compute the an ’s and the bn ’s?

Euclidean Spaces. Inner product
Let V be a vector space over R. An inner product or dot product is an
operation (•) between two elements of a vector space, satisfying the
following three properties: ~a
(1) Commutativity:
u~ • ~v = ~v • u~ for all u~, ~v 2 V .
(2) Linearity:
u~ • ( ~v + µ~ u • ~v ) + µ(~
w ) = (~ u•w
~) ~ 2 V.
for all u~, ~v , w
(3) Positive-definiteness:
For all u~ 2 V , u~ 6= ~0, u~ • u~ > 0.
Moreover,
u~ • u~ = 0 if and only if u~ = ~0.
Sometimes the dot product of u~, ~v is denoted as h~

u , ~v i. Examples.

Euclidean Spaces. Inner product Space
A vector space V over R furnished with an inner product •, and for that
we mean (V , R, •) is called a Euclidean vector space or an inner
product space.

Euclidean Spaces. Norm
Let V be a vector space over R. A norm is a map
k.k : V ! R, ~v 2 V ! k~v k.
satisfying the following three properties:

(1) Positivity:
k~v k > 0 for all ~v 2 V , ~v 6= ~0.
Moreover,
kv k = 0 if and only if ~v = ~0.
(2) k↵~v k = |↵|k~v k, for all ~v 2 V , ↵ 2 R.
(3) Triangle inequality:
k~v + w
~ k  k~v k + k~
w k, for all ~v , w 2 V .
Examples.

Euclidean Spaces. Normed space
A vector space V over R furnished with a norm k.k, and for that we
mean (V , R, k.k), is called a normed space.

Euclidean Spaces. Cauchy-Schwartz’s Inequality
Let V be a vector space over R. Then for all u~, ~v 2 V , the following
inequality holds:
p p
|~
u • ~v |  u~ • u~ ~v • ~v .

Euclidean Spaces
1 The notion of norm generalizes the idea of “length” for elements of

any Euclidean space.
2 If V is furnished with an inner product, then we can define the norm
p
k~v k = ~v • ~v .
3 (Cauchy-Schwartz’s inequality is needed to prove triangle

inequality). More common form of Cauchy-Schwartz’s inequality:
|~
u • ~v |  k~
u k k~v k.

Euclidean Spaces. Distance. Noise
A very important application of norms, is that they provide a notion of

distance for spaces where that idea is not intuitive.
In a normed space (V , k.k), given vectors u~, ~v 2 V we can define
u , ~v ) = k~
d(~ u ~v k.
So, for instance, for two signals, namely f (t) and g (t), such distance
between them is
s
Z b
d(f (t), g (t)) = (f (t) g (t))2 dt.
a
In particular, if f (t) is the signal that was sent, and g (t) the signal
received, then the above quantity is measuring the noise.

Orthogonality in Rn
In R2 and R3 , the angle between two vectors u~, ~v is defined as
u~ • ~v
cos(✓) = .
k~
u kk~v k
Furthermore, also in R2 and R3 , we say that u~, ~v are orthogonal (or

perpendicular) if ✓ = ⇡/2. In other words,
u~ ? ~v () ✓ = ⇡/2 () cos ✓ = 0.
Notice that whenever u~, ~v 6= ~0, this happens if and only if u~ • ~v = 0.
Notice that ✓ is the angle starting in u~ and ending in ~v .

E
Orthogonality in V
Given a Euclidean vector space V (therefore also furnished with the

corresponding norm) and u~, ~v 2 V , we define the angle between them
as the angle ✓ which fulfills
u~ • ~v
cos(✓) = .
k~
u kk~v k
Similarly, we say that u~, ~v 2 V are orthogonal if cos(✓) = 0.
u , ~v 2 V are orthogonal)
Whenever, this happens (~
D
u~ ? ~v () cos ✓ = 0 () u~ • ~v = 0.

Orthogonality in V
I
We say that a basis B of a Euclidean vector space V is orthogonal, if
every two vectors e~i , e~j 2 B, with i 6= j, satisfy that
e~i • e~j = 0.
Also, we say that B is orthonormal, if it is orthogonal and kei k = 1 for

all i.
Examples.
I
1123 V Space 1,011 CO1,07 estandar
innerproduct
Homework
Ip
XZ o
j fi o
Tf Xtyt 2 9 211,07 I
Joo
dft 2.1.07
L
Q
b MTT
Q prog e
w seize ii Tito IT
w
if ovz o
Xty12
0 grunted by 11 1,07 7
Leo 1 FI
IIIIIE
2 e
at
fl t
9 s
icand y
r NTI 2 147 1414 11µL 9
9
x 31 3 fu 9
I
d 413,2 20
14
dCR.IT d CE07 415 1207

74 7474 fE 3
free
IT d
b XtytZ 9 z
v z o Io't 90
x 1 a a team
Xm I 64
Hey27 0 9,0 411 2,1
East
2. Solve the following initial values problem where the initial value is given at t = 1:
(
u0 = Au II
type where A = 4 3 .
✓ ◆
u(1) = (8e4 5e, 5e)T 0 1
Alt A is Diagonalizable Hee

UCH e uld if
A PDF
Alt D Dp l
e pent
HI Al 4,71 6 1712 4
A
f 3 plat
TA 11,43
4,173
Le leer NI A Spae
p ri
i freecolumn I t D b Is
vz
Ker CHI Al span riots

D to 4 eDs feo
1 31
J u.in
tree n
TIpda.hinforsesdrr
et't unseat p
n.pe 7 pIop
Iuui
e
fi'oIl tEIEoe.oHo HlseIesef III 2

41 L
setsteEe4tI
oH
MENN
3, Solve the following initial values problem: type HI
k t
y 00 (t) y 0 (t)
0
2y(t) = et , y(0) = 1, y 0 (0) = 3.
yIH UeK
y't uz't U
et Udt 12447 et
y H7 U'zlH y1H 2gIt ya gilt qslt
uu i.fi u.f utetfiiiYaHiIInae etH ill feel
a
cat I'd
AsOBcs7ds
tucote.NO
eh ote
Solid ttltl
if A diagonalizable the fl
UCH Ee't ptuco Pe if ftpDsOE'Blsds
e
5 d 2 4 46 2
It path 1,1
L kerf I A e Sparkle DI
t
li
He.es
e am
2IffJeEeeEfLi
eAt.pefOtp yOfif1lEtEeI 3
Pe Dt
ask.ee 3 I'stain
eatuia.is e EiI
1S O S
eoezsH3 iIlEsIds f feoEsJfE

I e Bends.nl
0
as EI EIz L
it.in etiifl
e z 2
i e
d
Barrow's
rule
ilii
ee t.EE EIIEI7
in
Peet
et
Lte
ee e eIet i
Bluets
i
Have
un
t t
Et
yctl U.LA ez Eet
VERY
HARD II
0
4. Solve the following initial values problem: type
( 0 ✓ ◆
u (t) = Au(t) + B(t) 2 1
where A= , B(t) = (3e2t , 4e2t )T .
u(0) = (1, 1)T 0 2
I
plated 42 4
TA 2,23 Ma 2 2
Ker UI A Span HIM

A is Not DinoonhLIZABLE
In ie il
1 11 I I
Li.mu TAP spauKiio7il91Dff pJp
T
to 3 E to yordan Matrix
tee
r est Leite
IT F'AP A
Unltl Upttf fAEt B.tet c.EE

qt2e2t
Unt
yAze2ttBzte2t
sptncetiteztlhgpance.at te2tit2eU
Uld ffJ Aa 4 A t As
FOR JORDAN MATRICES CASE
if I 185 estate aeaei.tt
1 J
l eat aegtiee.FI
You can use this with no proof

B ert 2B.ee t2tCef
µ 2atIe 2Bzte2ttet
felt 2e2ttBze2t zezt2
2e'te't t itBatte't
Au a7EeY
t
Zest t 2Bzte2t 2cz
Blts t 24 29 a
2131 29 2131 132

q
282 2h72
I
typed✓
5. Solve the following initial values problem:
( 0 ◆
u (t) = Au(t) 1 1
where A= .
u(0) = (1, 1)T 1 3
t
plated 42 14
up
1 2 2
2
TA 12,23 ma
Lasker 21 A espouse MY
i'I in i uToi thetoriaruesonae
1411,11 173
Li Ias 2E APE Spout A PIF
T
881 fordantatix
Hit
E F'Antitheft'll fifth 974
L estate
Tf2eK
Inti uh.fieiIIIEIf fe7J
5
Orthogonality in V
Why are orthonormal bases useful?

Proposition
Let B = {~u1 , u~2 , . . . , u~n } be an orthonormal basis. Then the coordinates
of any vector ~x with respect to B, namely
(x1 , x2 , . . . , xn )B ,
can be obtained by simply multiplying ~x by u~1 , u~2 , . . . , u~n .
Proof.
Example.

Euclidean Spaces. Gram-Schmidt’s process
Gram-Schmidt’s orthogonalization process: given a non-orthogonal

basis B = {~u1 , u~2 , . . . , u~n }, we want to construct, from B, another basis
which is orthonormal.
Idea: At each step, construct a new vector e~j (starting from j = 1,

ending j at j = n), which is a linear combination of u~j and the j 1
previous e~j ’s, and which is orthogonal to e~1 , . . . , e~j 1 .
Algorithm (and example).

Orthogonality
The notion of orthogonality can be applied not only to vectors, but to
vector subspaces. The reason is the following result.
Lemma
Let {~v1 , ~v2 , . . . , ~vn } be vectors of V , and let
W = {~v 2 V such that ~v ? ~vi , for all i = 1, . . . , n}.
Then, W is a vector subspace of V .
(?) We will define W as the orthogonal subspace to ~v1 , ~v2 , . . . , ~vn
(?) One may see that in fact every vector in W is orthogonal to every
vector in Span(~v1 , ~v2 , . . . , ~vn ). So, we might write
W ? Span(~v1 , ~v2 , . . . , ~vn ).

Euclidean Spaces. Orthogonal Spaces
Two vector subspaces U and W are orthogonal, if every vector in U is

orthogonal to every vector in W , and conversely.

Euclidean Spaces. Orthogonal Complement
Given a vector subspace U, the space of all the vectors which are
orthogonal to U is called the orthogonal complement of U. It is
denoted by U ? .
Properties:
(1) If W = U ? , then
W? = U ) U ?? = U,
therefore dim U + dim U ? = n.

(2) Two subspaces can be orthogonal although one is not the
complement of the other (e.g. two lines in space).

Euclidean Spaces. Projection of a vector onto a subspace
Given a vector ~v and a vector Space V , the projection of v into V can be

computed following the algorithm:
1 We compute an orthonormal basis of V :
{!1 , !2 , · · · , !m }
2 The projection is
ProjV (~v ) = (~v • !1 )!1 + (~v • !2 )!2 + · · · + (~v • !m )!m .
Observe, moreover, that the distance between ~v and V is equal to
d(~v , V ) = kProjV ? (~v )k = k~v ProjV (~v )k.

z't
Leastminimumsquaresmethod ve
Tie UI E R ni ii are ei
A lui UI int eMkxmlR

µ
µ.aJ
rank A K
Normal equations
ATAZ AT b
Kx k text
ATA Atb
V is
The projection of 5 into
I
p A x
atix I A1ATAT'AT
Peg.ec imm
E B F E AE Cei emer
( 0
tepee✓ ◆
u (t) = Au(t) 2 1
where A= .
u(0) = (500, 100)T 1 4
t
p 5 61 9
3,33
TA Mats
II a
If E
II
a tap ftp.II.iit tf's Hit
P eat EE'T
eat iileatliiti.EEE.ie ILii

2He3t
2te3 2te3tC2t
ffl2 2 est
state
uay 6
type I 0 1
( 1 0 0
u0 (t) = Au(t)
A = @ 4 1 0A
U
where
u(0) = ( 1, 2, 30)T , 3 6 2
n
a
G Hkd 2
PHY
III Fat
Tess his
Math 1 Ma 1 2
it AP
L Kar CHI A Span1194613 j 42 Ker
NII
T 21 6 t 2
y
1
0 O O
I L I 12
19167 G O 21
Spae
En ai
ker 2E AI 01173
k Spar co
in it E fog
Iii JG O
a
A is NOT
DIAGONAL 17ABLE
7
4 O htet O
d P AP do
to
g
a est ft et ga
O O
O
et O
sente i n't ki
II see If I
t
t
EASY
8. Solve the following initial values problem: type I
( 0 1
u0 (t) = Au(t) 1 0 0
where A = @1 2 0 A.
u(0) = (1, 0, 0)T , 1 0 1
I
ai
a pCAH DU 216th
4 ker HI A
a
Spae 42 2,113 Ga 1,2 IS
since
lieI O 2
free Tiz
Ker 21 A Span o hoB L ter C II A spankoicia

u
E I O 3
3 E 3
I 3 a
Ei p
e't.feote.EE
O 0
o
solution Uche eatuca.pe F'III LEI ftp.go
t.io l tife oHItit I e 8

EASY
MEDIUM
0
( 0
type E
u (t) = Au(t) ✓ ◆
0 1
where A= .
u(0) = (2, 2)T , 1 0
t
ut PLA'tI d j te j
k
Lj Ker JI Al Span H1 IB
I
fIiIixtt ueu.p.fio.o
fIteEt l
ae
SpanLce9jI3
krC jI A
j I
Eff
sent
uneatuca it HIT Ift
it it
e i i jf.es e
feit e itljtfeitteotf
flesties't
tfest jest
jet ja
it cost 12 Sint
edt
422Sint
e
zjs.int
et't t e It 2cost
2 cost
fj I
j g1
9
( 0
u (t) = Au(t)
type's ✓ ◆
0 2
where A= .
u(0) = (1, 2)T , 1 2
T T
Phd 12 22 2
t
1 Hj d I j
t
The solution is
uctketfcostlnII sntf.tkD spawned by

11
111 I e'tas e't

Sint
10
11. Solve the following initial values problem: type III
( 0 ✓ ◆
u (t) = Au(t) + B(t) 2 5
where A= , B(t) = (2, 3)T .
u(0) = (1, 1)T , 1 2
t
f
Au
Solution UCH Ataco e
At
e Blu du
Platt 521 12 1 o
L f j
simples
A 1S L Span H jt2,173
Lj Kr JI Span
Jef
I
iii I
II it e fitz Itg
11
( 0 ✓ ◆
u (t) = Au(t) + B(t) 0 1
where A= , B(t) = (0, t)T .
u(0) = (7, 9)T , 1 0
12
tape I
3
y (t) + 2y(t) = 0a
( 000
y (t) 2y 00 (t) 0
e
y(0) = 3, y 0 (0) = 2, y 00 (0) = 6
III III uiu.io u ai.fi I

giYIIffftIIYgttIIykti rush that 24 t
zyu
A
at Ace
U'Itt U It a coil
T d l 0
PpfD
2
od
I if
d2
B 25 1 2 0
AH F 33 2
6 9 1,427 date ft NG 2
Uz
spanked K L A spanks HD's

L starlit A Kerth
in
a to
Us
hiker 2E At Span 1,2 4

T 13
a to to
fiiiz Solu
i
T I
u ia.eaeu
a.fi B
Homework i Solve y 2g t y get ateat
1 1,2
LINEAR ALGEBRA
November 11th, 2019
1. Given the matrix

0 1
1 2 a
A=@ 2 2 2a 1 A ,
3 10 3a + 1
where a 2 R.
(a) (1.5 points) Obtain the LU decomposition of A.
(b) (1.5 points) By using the previous LU decomposition, solve the linear
system AX = b, where bT = (4, 6, 16).
Solution:
(a) We are looking for matrices L lower triangular with main diagonal
entries equal to 1, and U upper triangular, so that A = LU , we get
such matrices after applying Gauss-Jordan method, obtaining:
0 1 0 10 1
1 2 a 1 0 0 1 2 a
13m51
A=@ 2 2 2a 1 A = LU = @ 2 1 0 A @ 0 2 1 A.
3 10 3a + 1 3 2 1 0 0 3
(b) We set Y = U X, then we need to solve first LY = (4, 6, 16)T ,

obtaining
b
0 10 yY
1 0 1 0 1 0 1
j
1 0 0 y1 4 y1 4
@ 2 1 0 A @ y 2 A = @ 6 A ) @ y2 A = @ 2 A .
3 2 1 y3 16 y3 0
Finally, we need to solve the linear system U X = Y ,

0 10 1 0 1 0 1 0 1
1 2 a x1 4 x1 2
@ 0 2 1 A @ x2 A = @ 2 A ) X = @ x2 A = @ 1 A .
0 0 3 x3 0 x3 0
u Y y
2. In the vector space of polynomials of degree at most 3 with real coefficients
T
we consider the bases
B1 = t, 1, t2 , t3
and t
B2 = t3 + t2 , t2 , t, 1 .
(a) (1.5 points) Obtain the matrix of change of bases from B1 to B2 , and
from B2 to B1 , identifying clearly which one is which.
(b) (1.5 points) Given the polynomial 1 t + t3 , use the proper matrix of
change of basis to calculate its coordinates with respect to B2 . Check
the result is correct.
Solution:
(a) There are many ways to solve this problem, we are going to use here
one of such methods. In order to obtain the change of basis from B1
to B2 we need to express each element of B1 with respect to the basis
B2 :
t t t t ttdd
o1 = 0(t
t = 0(t3 + t2 ) + 0(t2 ) + 1(t) + 0(1)
3
+ t2 ) + 0(t2 ) + 0(t) + 1(1)
!
!
u
t = (0, 0, 1, 0)B2 ,
1 = (0, 0, 0, 1)B2 ,
t2 = 0(t3 + t2 ) + 1(t2 ) + 0(t) + 0(1) ! t2 = (0, 1, 0, 0)B2 ,
and
t3 = 1(t3 + t2 ) 1(t2 ) + 0(t) + 0(1) ! t3 = (1, 1, 0, 0)B2 .
Therefore, the change of basis from B1 to B2 is
0 1
0 0 0 1
B 0 0 1 1 C
MB1 ,B2 = B C.
@ 1 0 0 0 A MBzibn ttbn.BZ
0 1 0 0
And to the the another case we have
0(t) + 0(1) + d
t3 + t2 = d E3 )
1(t2 ) + 1(t !
d f
t3 + t2 = (0, 0, 1, 1)B1 ,
t2 = 0(t) + T
0(1) + 1(t2 ) + 0(t3 ) ! tB ,r
t2 = (0, 0, 1, 0) 1
t = 1(t) + 0(1) + 0(t2 ) + 0(t3 ) ! t = (1, 0, 0, 0)B1 ,

and
1 = 0(t) + 1(1) + 0(t2 ) + 0(t3 ) ! 1 = (0, 1, 0, 0)B1 .
Therefore, the change of basis from B2 to B1 is
0 1
0 0 1 0
B 0 0 0 1 C
MB2 ,B1 = B C.
Bc 4711417
@ 1 1 0 0 A
1 0 0 0
1
Observe that MB2 ,B1 = (MB1 ,B2 ) .
0
IBD Mpsf MB.BE

MABE
Belt hit it
(b) Because we need to calculate the coordinates of 1 t+t3 with respect
to the bases B2 we need to express first the polynomial with respect
the bases B1 . In fact, we have
1I j t + t3t= ( 1,I 1,th

0, 1)fB ,
1
then the coordinates of the polynomial with respect to the basis B2

are
1 t + t3 = MB1 ,B2 ( 1, 1, 0, 1)T = (1, 1, 1, 1)B2 .
T
1 ttt
3. Let E be the space of squared matrices of dimension 2 with real entries.
We consider the linear mapping f : E ! E defined as follows:
f (A) = M A, Endomorphism
where in✓ a
◆
1 2
M= .
2 4
(a) (2 points) Obtain the coordinate matrix of f with respect to the basis
of E
B=
⇢✓
1 0
0 0
◆ ✓
,
1 1
0 0
◆ ✓
,
1 1
1 0
◆ ✓
,
1 1
1 1
◆
. Milf
(b) (1 point) By using the matrix obtained in (a), obtain f (C), where
✓ ◆
4 3
C= .
2 1
(c) (1 point) By using the matrix obtained in (a), obtain a basis of the
subspace ker(f ).
Solution:
(a) We need to express the image of each element from the basis into the
same basis, let us see how:
✓ ◆ ✓ ◆ ✓ ◆✓ ◆ ✓ ◆
1 0 1 0 1 2 1 0 1 0
f =M = = ,
0 0 0 0 2 4 0 0 2 0
and such matrix we need to express with respect to the basis B:

✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 1 0 1 1 1 1 1 1
=1 2 +2 +0 = (1, 2, 2, 0)B .
2 0 0 0 0 0 1 0 1 1
✓w ◆ ✓r ◆ ✓u ◆✓ ◆ ✓ u ◆
1 1 1 1 1 2 1 1 1 1
f =M = = ,
0 0 0 0 2 4 0 0 2 2
and
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 1 1 0 1 1 1 1 1 1
=0 1 +0 +2 = (0, 1, 0, 2)B .
2 2 0 0 0 0 1 0 1 1
✓ ◆ ✓ ◆ ✓ ◆✓ ◆ ✓ ◆
1 1 1 1 1 2 1 1 3 1
f =M = = ,
1 0 1 0 2 4 1 0 6 2
and
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
3 1 1 0 1 1 1 1 1 1
=2 5 +4 +2 = (2, 5, 4, 2)B .
6 2 0 0 0 0 1 0 1 1
For the last element we have
✓ ◆ ✓ ◆ ✓ ◆✓ ◆ ✓ ◆
1 1 1 1 1 2 1 1 3 3
f =M = = ,
1 1 1 1 2 4 1 1 6 6
and
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
3 3 1 0 1 1 1 1 1 1
=0 3 +0 +6 = (0, 3, 0, 6)B .
6 6 0 0 0 0 1 0 1 1
Hence, with all this information the coordinate matrix we are looking
for is 0 1
1 0 2 0
MBBo f
B 2 1 5 3 C
MB,B (f ) = B
@ 2
C.
0 4 0 A
0 2 2 6
T
(b) We express C with respect to the basis B, obtaining
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 1 1 1 1 1 1
C=1 +1 +1 +1 = (1, 1, 1, 1)B .
0 0 0 0 1 0 1 1
t T t t
Therefore ai ti it Ee
so, we get
O
f (C) ! MB,B (f )(1, 1, 1, 1)T = (3, 11, 6, 10)T ,
word of fCc
3uT Huizhou's ctoute ✓ ◆ w r t B
8 5
f (C) = (3, 11, 6, 10)B = .
16 10 µC
(c) The Kernel of the mapping is formed by the matrices A for whose
image is the null matrix, meaning:
⇢ E ✓0 0◆
ker(f ) = A : f (A) = .
EE
0 0
In other words, by using the previous matrix (remember this matrix

is with respect the basis B), we have
ker(f ) = (x, y, z, t)B : MB,B (f )(x, y, z, t)T = (0, 0, 0, 0)T .
From here we obtain the equations: Tat T

8
>
> x + 2z = 0 ⇢
<
II
2x y 5z 3t = 0 x + 2z = 0
)
>
> 2x + 4z = 0 y + z + 3t = 0
:
2y + 2t + 6z = 0
Therefore, a basis for the kernel (with respect to B) is:

⇣ ⌘
ker(f ) = Span ( 2, 1, 1, 0)B , (0, 3, 0, 1)B ,
I I a si on
so, writing the element with respect the original vector space, the
basis is E
✓✓ ◆ ✓ ◆◆
2 0 2 2
ker(f ) = Span , .
1 0 1 1
Il n
C OLB l9 B
TIME: 1 hour and 20 minutes.
HI a
p for
which A is diagonaizable
A tooo b D Dreguler
pro
pin.fi EIfE.fa
nHII.t hghliutf EI
a
T E
F Intl Iz o D
It
Ta I it rise i if
table
If Pto A is diagonal
2
Ta 94413 Mala 1 3 MgCheD

If f o
4 Karlie A
L
T
rameisIdn4IEuIaosa
o
A
t.EE
xty 0
42 tercee.at
b f O
L spank90117 Haid
fi Ivi ri 3
you
want Bjorn 0 i gFz
o
83 Lacy z then
L
Z 0
atyeo TI
V3 Ct
El It t'm
p it
ti io t L to
et.FI tII
Sheet 7 Euclidean Spaces
d but not orthogonal each

1 Gramm Schmidt U U w i
other
U v to o
has
no
ieiv.ru
s t they are or the
22
v3 Sparegun 83
step i b
re
Second i span
it ie IV Ito
O E Te Fr Av I re Av re noon dIo
a if von o I V
D I
IaIY
u
Alf v u to
IET us
UT u r w3 qan u in
third Spain
I Te fair yw w I O
Wo I O
n
µ re u
a
o
yw a u.uipeu.az
r
M
a uw
o_u yw
L
o
o wit Cu Mu na v
yw up go.eu
o
Y u.o
ua.otuni.uv.o rw.ir
iuZI.vo r j
u.vCi u1
w
Department of Physics and Mathematics LINEAR ALGEBRA (350000) Course 2019/20
Euclidean Spaces
1. Find, using Gram-Schmidt orthogonalization, an orthonormal basis of the subspace of R4

spanned by a1 = (2, 0, 2, 0), a2 = (1, 1, 1, 0), a3 = (2, 2, 0, 0).
A 2,012,0
92 11 1,107
93 12 40,07
We define it s t Space Vi span at then iiA no ia
Tin
Next we consider Tiz Spaulat Span Tinted and iii
then
Th ta 11,0 no 111 1,40 since I.ee e o
Uz
0 2 d 2 d l Fez 0,1 0,0
Spau aI.difad spancih.UIUD.u.us 0 an

is s t
Finally UJ
Plz Tiz o We consider
0 then
since Teno
UT h l O o
µ
UT
YUI
then
112 and became UT
Us o
0 1 fu 2 µ
t
y y l therforei Tt's
1,0 1210 so
we can instead take Us 1,0 1,0

1
2. Consider the vector subspace U = Span{(1, 0, 1)} of R3 , with the usual dot product.
Find the implicit equations of the orthogonal complement U ? . Also, find the orthogonal
projection of (1, 1, 2) onto U and U ? , respectively.
Ut X y HE
1123
CXy Z no e o
X y HE1123 x z o Span 1 07 t
f Ei
we solve the implicit egnatim
o CIePgaEYfn
r
Ll O
pivot free
XtZ
t
colures
jxyz.jp of Ut
Because Th Tiz o then
Proj Hit27 H
uh y uY
uit
tuff Ba flat in Ethan l E i E
so since
proju 1,427 project 1,12 442 then
projuu1,27 4 e 2 projectit 2 Z o
it is also equal
to r
1412 of v
To Tr
2
3. Consider the vector subspace U of R3 , furnished with the usual dot product, defined by
U ⌘ x + y = 0. Find the parametric and implicit equations of the orthogonal complement
U ? , and the orthogonal projection of b = (1, 2, 3) onto U .
spankomD It.io
Pe Uz
u
oTffgIdd 2 O
Ut i XyZ7ElR3
Hey c
1123
Kiya
xty3
s
1010,17 0 and Hey Z I 1,407 03
Span K 1073
as
d mania
i.sIuuIuTs
proiulhzid.lt ffi.o9Yifio 4ind Ilana Zo
projultih37 142,37 projuelhad fE E 3
3
4. In R4 , with the usual dot product, consider the vector subspace
U = L ((1, 1, 1, 1), (1, 1, 1, 1), (1, 1, 1, 1)) .
(a) Find a basis and the implicit equations of the orthogonal complement U ? of U .
(b) Construct an orthonormal basis for U .
(c) Express v = (1, 2, 3, 4) as a sum of two vectors, one of them belonging to U and the
other to U ? .
(d) Find the distance from v to U and U ? .
5. Find an orthogonal basis of the subspace U ✓ R4 of equation x+y z +w = 0. Determine

the projection of v = (1, 0, 0, 1) onto U and the distance from v to U .
6. As in problem 3, determine the orthogonal projection of b = (1, 2, 3) onto the subspace

spanned by the columns of the matrix
0 1
1 0
A = @ 1 0 A.
0 1
(a) Solve the normal equations
AT Ax = AT b,
and after finding the solution x of this system, determine the projection as p = Ax.
2
4 Us Span It 1 htt Hilda It 1 1 11
in tis
a Ut din U 3 dim Ut L
Implicit egs
t of UL
t o t o X y z t o
U X
ytz xtytz
t t
E OI OIT2 Od Iz outspanHoi hah
II
o
t
ye d
free
4 0
column param
egs
b VI Il I l l became UT Tiz o JI UI CeeAD
Tsui KEITH It hi D Ela na D Fine eat

d LTE 1 1 1 1 I 1 1 1 1,0 1 o
wi
II L ftp t Hzit.Ii II
11h11 It FEZ
ri.IE
rtTtru'EiEiliyz.E.EiE w3
vI l1iQ o
IT 9 10
Vdl f p
4. In R4 , with the usual dot product, consider the vector subspace
U = L ((1, 1, 1, 1), (1, 1, 1, 1), (1, 1, 1, 1)) .
(a) Find a basis and the implicit equations of the orthogonal complement U ? of U .
(b) Construct an orthonormal basis for U .
(c) Express v = (1, 2, 3, 4) as a sum of two vectors, one of them belonging to U and the
other to U ? .
(d) Find the distance from v to U and U ? .
work Solvethe same problem without vector 411,1 in U

1 n i with the inner product
xcy2ctloolxiycz.II xxtzyy 32E 4LI
4
we have ai Chi tch M Ue Obtain orthogonal basis
Tri le 1 1 e it
UI 1,1 0,07 272
Tiye 110,0 1 Tis
problem
can rename them to simplify the
you
Vi U7 lh h4
1St
CUT Span If U and oiortz
vz.ua
21T Spar
Joi I.ro Vo drToPi 2 0

11
Ho o 4
By
definition
It 5
31T spae VI Fz B Span Un UzUs and vi too
by
def Tz Vz o
vii vis wi tr to
r
VI 83 of prior
8 OUT
0 0 d4 o
07
12
Ijv3 vIou5 dvz.v7 peFz.B 0 1 od p2
r
Lg defiant
fidef
by
UT 11,010,1 Vj 11,0 o D O VI 1,1010
ki Iz o t 83 1 110,2
vii FatsFy Te TE'sUT and Fiora 0
1st Span Spon
VET Try to
vii nie Toi peut y VI Tie o
tofu TrioUI d 07ooh K 52 qrf.ir

VoUys d 8Br fUVorz qFoI
VoH
E oui ti tea airs
2 112
2 d 4
2 o
0 0 In µ
6 0
0 0 y y
If Ii Evi h h t D Ell 1,1 D
E I Z E
By ft 1 3 l
5. Find an orthogonal basis of the subspace U ✓ R4 of equation x+y z +w = 0. Determine
the projection of v = (1, 0, 0, 1) onto U and the distance from v to U .
5
6. As in problem 3, determine the orthogonal projection of b = (1, 2, 3) onto the subspace
spanned by the columns of the matrix
0 1
1 0
A = @ 1 0 A. a o
0 1
(a) Solve the normal equations

AT Ax = AT b,
and after finding the solution x of this system, determine the projection as p = Ax.
(b) Find the projection matrix
P = A(AT A) 1 AT ,
and check that p = P b is the projection computed in (a).
Ate ff 9 ATA 9 Atan f Normal

at
tool
ui
F AI
f 1 14 1 e b pelts fi
E Alata'At I 8 8911119
it H L
b b
p P
6
e
p
7. Find the regression line b = CL+ Dt L fort he data t1 = 1, t2 = 2, t3 = 3, t4 = 4, b1 = 3,
b2 = 4, b3 = 5, b4 = 7 (i.e. the line of equation b = C + Dt that best approximates, in the
least squares sense, the points (1, 3), (2, 4), (3, 5), (4, 7)). Denoting the solution p = Ab
(the projection of b onto the column space of A, which provides the values at the points
ti of the computed regression line), explain why the following relationships, well-known
in statistics, hold:
e1 + e2 + e3 + e4 = 0, t1 e1 + t2 e2 + t3 e3 + t4 e4 = 0.
t
Ht t.tt I H n
7
T
T A X
b
I P R b A ATA Atb
a
Ii e Iii's.li I iHe I 3
l ni I ai
t
elite.l t lt
sat
i
e b r HH H 7
8. Consider the problem of calculating a regression parabola b = Dt + Et2 (i.e. we impose
that the constant term is C = 0) for the data t1 = 1, t2 = 2, t3 = 3, t4 = 4. Let x be the
solution vector of the normal equations AT Ax = AT b and let e = b Ax tbe he residues
vector. Explain why (regardless of the vector b) it holds that
e1 + 2e2 + 3e3 + 4e4 = 0,
although in this case we cannot ensure that the sum of all the residues
e1 + e2 + e3 + e4 = 0.
is zero. Deduce other relationship that is also fulfilled by e1 , e2 , e3 , e4 .
t il by
it HEiitn
Hence
youmustuse
i
least min sopreres method
it I 9
youmustfind
I
A
f If
ATA
ATA l E f Atb A
t the colours of
II
is orthogaf.ae w r
f SI me
b AI
t t e tzez tz.es t tyg
0
I
3 I tie tf ez ts est TI 4
0
620 8
E 3 05 bi b e2bz 134 4by

620 b 14be 963 16bn
e L
H iiI LINEAR ALGEBRA
Third Evaluation Test
i
January, 16th, 2019
tT
1. In R4 with the standard scalar product, we consider the subspace :
Io4zbI4
U = h( 1, 1, 0, 1), (1, 1, 2, 1), ( 3, 1, 0, 1)i .
test 3es
b
e
41zbfytfzb.jo iguf7lb
Se pide:
2bat3b3t4h
a) Obtain a basis of the orthogonal complement U ? of U . (1 pt.)
620
b) Compute the orthogonal projection of the vector b = (2, 5, 1, 1)
onto the subspace U . (1,5 pts.)
20 2.100 3.240 4.480
c) Compute the distance between b and U . (1 pt.) f b t 4b z t 9b t 16by
6 20 the definition of U ? , then
Solution: a) Taking into account
U ? = Span (1, 2, 1, 1) .
0
b) Taking into account projU (b)+proj U ? (b) = b, we have since
0 1 0 1
1 2
B 2 C1 B 4 C
projU ? (b) = A(AT A) 1 AT b = B C B
@ 1 A 7 14 = @ 2 A ,
C
254 where A = (1, 2,

1 2
6I6 1, 1)T , so
0 1
0
48556 B 1 C
projU (b) = b projU ? (b) = B C.
7316
@ 1 A
1
c) An the distance is
5620 d(b, U ) = kprojU ? (b)k =
p p p
22 + 42 + ( 2)2 + 22 = 28 = 2 7.
2. For the following data:
x1 = 1, x2 = 0, x3 = 2, x4 = 3,
y1 = 0, y2 = 1, y3 = 3, y4 = 2.
We want:
a) To compute the regression line y = a + bx (i.e. the line y = a + bx
that fits the best the the points (xi , yi ) in the sense of least squared).
(2 pts.)
b) Evaluate the line of a) in the values xi . (0,5 pts.)
c) If e = (e1 , e2 , e3 , e4 ) is the residue vector, explain why, in this case,
the relación e2 + 3e3 + 4e4 = 0 fulfills. (1 pt.)
Solution: a) If we want to obtain such regression line we have with these
values 0 1 0 1
0 1 1 ✓ ◆
B 1 C B 1 0 C
B C B C a .
@ 3 A=@ 1 2 A b
2 1 3
To get the solution of the regression line, y = a + bx, we need to solve the
linear system ✓ ◆ ✓ ◆✓ ◆
6 4 4 a
= .
12 4 14 b
We get as solution:
9 6
a= , b= ) 10y = 9 + 6x.
10 10
b) With this solution we have
3 9 21 27
y(x1 ) = , y(x2 ) = , y(x3 ) = , y(x4 ) = .
10 10 10 10
c) In this case we have
3 1 9 7
e 1 = y1 y(x1 ) = , e 2 = y2 y(x2 ) = , e3 = , e4 = .
10 10 10 10
Since ~e is orthogonal with the matrix A, i.e.
e1 + e2 + e3 + e4 = 0, e1 + 2e3 + 3e4 = 0,
then the condition we have is the sum of the two columns of such matrix,
3. Given the boolean function
f (!, x, y, z) = (xz + !)y + x,
we want to compute, by using its value table, its disjunctive normal form
and its conjunctive normal form. (3 pts.)
Solution: The table of values of f is
! x y z xzy !y I
n f i wxEow1y
0 0 0 0 0 0 0 0
µ
0 0 0 1 0 0 1 0 o O O
0 0 1 0 0 0 2 0 O o o
0 0 1 1 0 0 3 0 co O 0
0 1 0 0 0 0 4 1
1 1 O
0 1 0 1 0 0 5 1
O 0 O
0
DO
1 1 0 1 0 6 1
toO
0
1
1
0
1
0
1
0
0
0
0
0
7
8
1
0 IO no
1
1
1
0
0
0
1
1
0
0
0
0
1
9
10
0
1
O l O
0 1 l
1 0 1 1 0 1 11 1
E a I
1 1 0 0 0 0 12 1
1 y O
1 1 0 1 0 0 13 1
O O
g l
1 1 1 0 1 1 14 1
t n
1 1 1 1 0 1 15 1 O 1 I
Therefore the d.n.f. of f is
f (!, x, y, z) =
X
u10, 11, 12,
u u6, 7,
m(4, 5, u 14,
u 13, u = !xyz + !xyz + !xyz + !xyz + !xyz
u 15)
+ ! x y z + ! x y z + ! x y z + ! x y z + ! x y z.
and the c.n.f. of f is

Y
f (!, x, y, z) = M (0, 1, 2, 3, 8, 9) = (! + x + y + z)(! + x + y + z)(! + x + y + z)(! + x + y + z)
(! + x + y + z)(! + x + y + z).
LINEAR ALGEBRA
January, 16th, 2019
U = h( 1, 1, 0, 1), (1, 1, 2, 1), ( 3, 1, 0, 1)i .
Se pide:
ii i H
onto the subspace U . (1,5 pts.)
ei
c) Compute the distance between b and U . (1 pt.)
to Ii
Solution: a) Taking into account the definition of U ? , then i
U ? = Span (1, 2, 1, 1) . free
coburn
b) Taking into account projU (b)+proj U ? (b) = b, we have since
0 1 0 1 t A
I
B 2 C1
1
B 4 C
2 2 X
projU ? (b) = A(AT A) 1 AT b = B C B
@ 1 A 7 14 = @ 2 A ,
C
1 2
where A = (1, 2, 1, 1)T , so

0 1
0
B 1 C
projU (b) = b projU ? (b) = B
@
C.
1 A
1
c) An the distance is
p p p
d(b, U ) = kprojU ? (b)k = 22 + 42 + ( 2)2 + 22 = 28 = 2 7.
x1 = 1, x2 = 0, x3 = 2, x4 = 3,
y1 = 0, y2 = 1, y3 = 3, y4 = 2.
We want:
a) To compute the regression line y = a + bx (i.e. the line y = a + bx
(2 pts.)
b) Evaluate the line of a) in the values xi . (0,5 pts.)
the relación e2 + 3e3 + 4e4 = 0 fulfills. (1 pt.)
Solution: a) If we want to obtain such regression line we have with these
values 0 1 0 1
0 1 1 ✓ ◆
B 1 C B 1 0 C
B C B C a .
@ 3 A=@ 1 2 A b
2 1 3
To get the solution of the regression line, y = a + bx, we need to solve the
linear system ✓ ◆ ✓ ◆✓ ◆
6 4 4 a
= .
12 4 14 b
We get as solution:
9 6
a= , b= ) 10y = 9 + 6x.
10 10
b) With this solution we have
3 9 21 27
y(x1 ) = , y(x2 ) = , y(x3 ) = , y(x4 ) = .
10 10 10 10
c) In this case we have
3 1 9 7
e 1 = y1 y(x1 ) = , e 2 = y2 y(x2 ) = , e3 = , e4 = .
10 10 10 10
Since ~e is orthogonal with the matrix A, i.e.
e1 + e2 + e3 + e4 = 0, e1 + 2e3 + 3e4 = 0,
then the condition we have is the sum of the two columns of such matrix,
f (!, x, y, z) = (xz + !)y + x,
we want to compute, by using its value table, its disjunctive normal form
and its conjunctive normal form. (3 pts.)
Solution: The table of values of f is
! x y z xzy !y n f
0 0 0 0 0 0 0 0
0 0 0 1 0 0 1 0
0 0 1 0 0 0 2 0
0 0 1 1 0 0 3 0
0 1 0 0 0 0 4 1
0 1 0 1 0 0 5 1
0 1 1 0 1 0 6 1
0 1 1 1 0 0 7 1
1 0 0 0 0 0 8 0
1 0 0 1 0 0 9 0
1 0 1 0 0 1 10 1
1 0 1 1 0 1 11 1
1 1 0 0 0 0 12 1
1 1 0 1 0 0 13 1
1 1 1 0 1 1 14 1
1 1 1 1 0 1 15 1
Therefore the d.n.f. of f is

X
f (!, x, y, z) = m(4, 5, 6, 7, 10, 11, 12, 13, 14, 15) = ! x y z + ! x y z + ! x y z + ! x y z + ! x y z
+ ! x y z + ! x y z + ! x y z + ! x y z + ! x y z.
and the c.n.f. of f is

Y
f (!, x, y, z) = M (0, 1, 2, 3, 8, 9) = (! + x + y + z)(! + x + y + z)(! + x + y + z)(! + x + y + z)
(! + x + y + z)(! + x + y + z).
10 fWXy t Em 0,213,4167,819 11,12 13,15 Ndrf
GCW X y t Emts 416,719 11,121314,15 dnf
a Table of
values
of f g and fg
b dnf off Cnf of g
eo oot8of.g
l O
O O 0
f t
O O utxy z.ir Wtxtytz
1 z w x y z.ir
wtXtytI
I
I
8h19 f ft
81 o
9 A 1
001 10 O O
f
Il o
1
7
8
r
G I'WIFFEN
tw xyz.ir
i
Etxtytz
Cotxtytz
l g g twxyz.ir
1011
1 to f f i lo twx y z.ir
O r y twxyZ
1Y t
lionO 04
1 101 at wxyz
fake
y XI II Is
wzxtwz wzy
dkrarnaughiwz.rs
WE
wEf
w
E
II
Z2Dp sIyItIyw
txyzfxgtwtfxy
wy
xyw
z way
ITw xy.az
f wiki4,7 WZXtWZ ETtw w tx 4I x y
ozytxyw xyz fxywtxy I.IT wtIy E

h h
F
or
w x
yizl WTtwT YZ y
hnthytyItwz
helw.ly
2 w
LINEAR ALGEBRA
retrieval of the First Evaluation Test
January 17th, 2020
1. Given the matrix 0 1

1 0 1 1
00
B 2 2 1 2 C
A=B
@
C
1 2 3 0 A
2 0 1 6
a) Obtain the LU decomposition of A, i.e. A = LU . (1’5 pts.)

b) Using the previous result to solve the linear system Ax = b, where
b = ( 1, 5, 4, 7)T . (1’5 pts.)
d 00 d 000 I O ty
a I o sy
2 1 00 O u uzUg 2 1 00 O 2 10
O O UyUs I A0 O O Uy Us
1 e AO 0 O O
2h2 l3 A 0 On
go 2h2 last Ut
U 2 2l 2 lr
2 62 1
L 1 11 44 3 s 44 9
Zt Uz
1 145 0 Us I
1 000 I O l 1
2 1 00 O 2 IO
O 01 I LU
l l 1O
O O 03
Zo 7 A
b At b L b UX Y D Ly b
Ux
1 00
2
Y
Ii iiit.ftxi.it
2. In the vector space R3 we consider the bases
ill in its
B1 = {(1, 2, 3), (3, 1, 1), ( 4, 1, 0)}
and
B2 = {(1, 0, 1), (3, 0, 0), (0, 1, 1)} .
Let P the matrix of the change of bases from B1 to B2 , and Q the matrix
of the change of bases from B2 to B1 .
a) Obtain matrices P and Q. (2 pts.)

b) Given the vector (4, 1, 2) 2 R3 , by using the matrix obtained in a),
obtain the coordinates of such vector with respect to the basis B2 .
(1 pt.)
I
a MB Be
a
MB7132 MBC BIMBABc MpoBz 12g I
i
z 1 I
a F the 2 3 1
a t
b 4 12 413,8 B E 1,1 1 Be
11 Iz Iz Iz
L U t file 843
3. Let E be the vector space of 2-by-2 matrices with real entries, and let F
be the vector space of polynomials of degree less or equal to 2 with real
coefficients, and let f : E ! F be the linear mapping defined by:
✓✓ ◆◆
a b
f = (a + b)x2 + cx + d,
c d
a) Compute the matrix of the application with respect to the basis:

⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 1 1 0 1 0 1
B1 = , , ,
0 0 1 0 1 0 0 1
A Ar
of E and B2 = {1, 1 x, x2 } of F . (2 pts.)As Ay
✓ ◆
1 2 t A
b) Let M = . Find f (M ) by using the matrix obtained in
0 1
a). (1 pt.)
c) By using the previous matrix obtained in a), compute a basis and
the dimension of the kernel of f , ker(f ). (1 pt.)
Exam duration: 1h 20 minutes
a f Aa x2 010117132
IIE
f Az 2X2tX A 11432
msn.it
i'iii I
b f IfII MB.at IEy l no 37Bz 3
2 1
f E Z'Z 1
B g
wi t B1
iii iii iii

iii I tt L
i H tt 1 i ii
LINEAR ALGEBRA
Second Evaluation Test
January 17th, 2020
1. Given the matrix:

0 1
1 b+1 2
A=@ 0 b 0 A,
4 1 1
with b 2 R.
a) Determine the values of b for which A is diagonalizable over R. (3
pts.)
b) For b = 1. Obtain the matrices D, diagonal, and P , invertible, so
that A = P DP 1 . (2 pts.)
a We compute the characteristic polynomial of A

an bt 2 al 2
Det AI A 0 0 ab la b 12 I 8 2 b 2 3 213
4 Atl
4 1 Att Twc
expandthedetalong
the 2ndrow
T A b 3 33 So i
spectrum of A is
therefore the
A is diagonalizable
a if bt 3 3 then A has 3 diff eigenvalues
3i
then 137 2 we need to compute Mgl
If b 3 Ma
Ma 137 1
23 AI A ramie 3D A 2 Mg 37 7 Ma 3
3 then Ma137 2 we need to compute Mgl3

If b
Ma 3 f
4 2
02 Mg137 7 Ma63
2 3 AI A
fo g 2
ram RBI A 2
b if we set now b t them l Z 2
A o t O
4 t t
en
L
A as
and Kerl 3
Ui
L.kz
SI a L II L.la a 211
itI L3
I A
E
Ker 3II A
23
116,47
D Then I ai ii iiiIf.dz 4
3I A
2. We consider the initial value problem
✓ ◆
1
X 0 = AX + G, X(0) = ,
4
where ✓ ◆
1 1
A=
2 2
and ✓ ◆
et + 3
G(t) = .
et 6
a) Compute the matrix eAt . (2 pts.)

b) Solve the initial value problem by using the matrix eAt . (3 pts.)
Exam duration: 1h 30 minutes.
Ca In order to obtaineatwe need to obtain D I so that A PDP then
eat I Dtp
Inthis case INI Al AI Iz I 32 0 HA 0,3 There D Oo
and to HerCott A A
f z
Lo Lkl N
13 14131A 31 A
f say 2g
E
f I
eat.fi.ME EIkitsIieIeeIitt.e.L EIi EI

at t o then the solution
b Since X_AX G andthe initial condition is
is
µH eatXD e.AToteAsGcs ds
oeAtXlol f L2tze7Iffzee7e ffJ f2 Yet X Homogeneous solution

Teas asus tf
Lee's as
I as
L II IEEE then
I 2
eatteasaads.es EEIiIIIIEEaeIe.sf fEIEe sf xpie
particular
solution
Then the solution is
xlth xniti.ae kiEIfEIzeEEIffE.eeI

LINEAR ALGEBRA
January 17th, 2020
U = h(1, 1, 2, 0), ( 1, 3, 2, 1), (2, 0, 1, 2)i .

onto the subspace U ? . (1,5 pts.)
c) Compute the distance between b and U ? . (1 pt.)
a Ut X yZ t X y122 0 x13L 221 2 0 2 2 22 03 Now we
f H H p.mn
Ut Lf 3 1,2 2
b Inthis case projut B Bol3,4227 314 2

13h221.63e
If 31,3 2 13 Li t l
c Since d Brut Hproju15711 andtakinginto accountthat

D proju B projuelII then
B l tZ
proju B projut Ii 1,1
DCB Ut ftp.fff fz
3
x1 = 0, x2 = 1, x3 = 2, x4 = 3,
y1 = 1, y2 = 0, y3 = 2, y4 = 2.
We want:
a) To compute the regression line y = C + Dx (i.e. the line y = C + Dx
(2 pts.)
b) Evaluate the line obtained in a) in the values xi . (0,5 pts.)
the relation e1 + 2e2 + 3e3 + 4e4 = 0 fulfills. (1 pt.)
need to solve is
a The linearsystem we
ea
i ti
I t.tl
A b
1
AtA If
in
d
Atb in this case
regular matrix
and 5 then the regression

f E o
e
f
is 11
line y 10
b Fr Yo yT yet Fu 34
0
512
t o
910 izeztses 4e 4tzeo
He 5 444
1 fI jo fIIgogffej1ei
f (!, x, y, z) = (y + x z)(z + x y),
we want to compute, by using its table of values, its disjunctive normal

form and its conjunctive normal form. (3 pts.)
Exam duration: 1h 30 minutes

wxy
W 0123 4567 891011 12131415

fiflwix.ly z dIm t
sum
112,317,910,1415
d
WITZ
w 0000 0000 1111 1111 g
X 00001111 00001111 cnf flw y TIM 014,516,812,1314
0011
r L
y O 11 00110011 product
wt y z
01010101 0101 of wtxi.ge
Z 0101
Iz 0101 00000101 0000
ytIz 0111 0011 0111 0011
Is 0011 0000 0011 0000
2 Is 0111 01010111 0101
f 0111 0001 0111 0001

21.notes - Linear Algebra

Uploaded by

Document Information

Original Title

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

21.notes - Linear Algebra

Uploaded by

Copyright:

Available Formats

Academic Year 2020-21 Linear Algebra

Roberto Costas (E223)

Bibliography LAY D e linear Algebra and its apps 14thEd

other material BB www.rscosan.com docencia.htm

Exams PEI PETE PEI 1 31 1131113

3019 W 10 00 12 00 in room SA5B 22515 Ll t

September 23, 2020

Matrices, Determinants and Linear Systems

A matrix Am⇥n is an array of numbers in rows and columns

We say that the dimension of A is m ⇥ n (we also could say m-by-n). If

Matrices, Determinants and Linear Systems

Main diagonal. If A is a square matrix of dimension n, the elements aii ,

Matrices, Determinants and Linear Systems

Operations: Khan Academy

Sometimes we will write

Matrices, Determinants and Linear Systems page5 I

3. Multiplication of two matrices. Properties:

Matrices, Determinants and Linear Systems page 6

Matrices, Determinants and Linear Systems

wher L is a lowertriangular matrix

This is the so called LV decomposition

The Gauvin method let us to

it is non zero entry

so det Ae AnAzz Arazi Infact pageG

In fact if A is uppertriangular deLCA U Un Um

Matrices, Determinants and Linear Systems page o

Matrices, Determinants and Linear Systems page1

Matrices, Determinants and Linear Systems page1

Computation of the inverse of an square matrix A.

Matrices, Determinants and Linear Systems 1

We say that a certain row r (similarly for columns) is a linear

for certain numbers ↵1 , . . . , ↵s , which are called the coefficients of

Matrices, Determinants and Linear Systems page1

An equivalent definition of rank, in terms of determinants. A minor,

Matrices, Determinants and Linear Systems page1

Matrices, Determinants and Linear Systems page 1

Some rules for computing rank(A):

Matrices, Determinants and Linear Systems page

A System of Linear Equations is a set of equations of the type

Matrices, Determinants and Linear Systems page1

The system can be written in matrix form in the following way:

Question: why is the above equality true?

Matrices, Determinants and Linear Systems page1

Classification of Linear Systems: linear systems can be

Matrices, Determinants and Linear Systems page19

In order to classify a given linear system, we use the augmented matrix,

Matrices, Determinants and Linear Systems page2

Theorem (Rouche-Fröbenius Theorem)

When rank(A) = rank(B) = n, the di↵erence n rank(A) is in fact the

Matrices, Determinants and Linear Systems page2

Matrices, Determinants and Linear Systems page2

Always consistent (why?)

Matrices, Determinants and Linear Systems page2

Azz 2 L Ucz l2Uktl3Urz 313 13 3

Lesson 2: Vector Spaces n r 7 2,0 1

n the same space R2 vectors

(i) (V , +) is a conmutative group:

Lesson 2: Vector Spaces page

(ii) The operation · satisfies that:

Lesson 2: Vector Spaces page2

In order to make explicit whether we work over R or C, one writes

Inthe previous example the space of all

Two elements of V are