MASSACHUSETTS INSTITUTE OF TECHNOLOGY

6.436J/15.085J Fall 2018

Problem Set 1

Readings:
(a) Notes from Lecture 1.
(b) Handout on background material on sets and real analysis (Recitation 1).
Supplementary readings:
[C], Sections 1.1-1.4.
[GS], Sections 1.1-1.3.
[W], Sections 1.0-1.5, 1.9.

Exercise 1.

(a) Let N be the set of positive integers. A function f : N → {0, 1} is said to be

periodic if there exists some N such that f (n + N ) = f (n), for all n ∈ N.
Show that the set of periodic functions is countable.

(b) Does the result from part (a) remain valid if we consider rational-valued
periodic functions f : N → Q?

Solution:

(a) For a given positive integer N , let AN denote the set of periodic functions
with a period of N . For a given N , since the sequence, f (1), · · · , f (N ),
actually defines a periodic function in AN , we have that each AN contains
2N elements. For example, for N = 2, there are four functions in the set
A2 :

f (1)f (2)f (3)f (4) · · · = 0000 · · · ; 1111 · · · ; 0101 · · · ; 1010 · · · .

The set of periodic functions from N to {0, 1}, A, can be written as,
∞
!
A= AN .
N =1

Since the union of countably many finite sets is countable, we conclude that
the set of periodic functions from N to {0, 1} is countable.

(b) Still, for a given positive integer N , let AN denote the set of periodic func-
tions with a period N . For a given N , since the sequence, f (1), · · · , f (N ),

1
 actually defines a periodic function in AN , we conclude that AN has the
same cardinality as QN (the Cartesian product of N sets of rational num-
bers). Since Q is countable, and the Cartesian product of finitely many
countable sets is countable, we know that AN is countable, for any given N .
Since the set of periodic functions from N to Q is the union of A1 , A2 , · · · ,
it is countable, because the union of countably many countable sets is count-
able.

Exercise 2. Let {xn } and {yn } be real sequences that converge to x and y,
respectively. Provide a formal proof of the fact that xn + yn converges to x + y.
Solution: Fix some ϵ > 0. Let n1 be such that |xn − x| < ϵ/2, for all n > n1 .
Let n2 be such that |yn − y| < ϵ/2, for all n > n2 . Let n0 = max{n1 , n2 }.
Then, for all n > n0 , we have
ϵ ϵ
|(xn + yn ) − (x + y)| ≤ |xn − x| + |yn − y| ≤ + = ϵ,
2 2
which proves the desired result.
Exercise 3. We are given a function f : A × B → R, where A and B are
nonempty sets.
(a) Assuming that the sets A and B are finite, show that

max min f (x, y) ≤ min max f (x, y).

x∈A y∈B y∈B x∈A

(b) For general nonempty sets (not necessarily finite), show that

sup inf f (x, y) ≤ inf sup f (x, y).

x∈A y∈B y∈B x∈A

Solution:
(a) The proof rests on the application of the following simple fact: if h(z) ≤
g(z) for all z in some finite set Z, then

min h(z) ≤ min g(z) (1)

z∈Z z∈Z
max h(z) ≤ max g(z). (2)
z∈Z z∈Z

Observe that for all x, y,

f (x, y) ≤ max f (x, y),

x∈A

2
 and Eq. (1) implies that for each x,

min f (x, y) ≤ min max f (x, y).

y∈B y∈B x∈A

Now applying Eq. (2), let’s take a maximum of both sides with respect to
x ∈ A. Since the right-hand side is a number, it remains unchanged:

max min f (x, y) ≤ min max f (x, y),

x∈A y∈B y∈B x∈A

which is what we needed to show.

(b) Along the same lines, we have the fact that if h(z) ≤ g(z) for all z ∈ Z,

inf h(z) ≤ inf g(z) (3)

z∈Z z∈Z
sup h(z) ≤ sup g(z). (4)
z∈Z z∈Z

These follow immediately from the definitions of sup and inf.

As before, we begin with

f (x, y) ≤ sup f (x, y),

x∈A

for all x, y. By Eq. (3), for each x,

inf f (x, y) ≤ inf sup f (x, y),

y∈B y∈B x∈A

and using Eq. (4),

sup inf f (x, y) ≤ inf sup f (x, y).

x∈A y∈B y∈B x∈A

Exercise 4. A probabilistic experiment involves an infinite sequence of trials.

For k = 1, 2, . . ., let Ak be the event that the kth trial was a success. Write down
a set-theoretic expression that describes the following event:

B: For every k there exists an ℓ such that trials kℓ and kℓ2 were both successes.
" #
Note: A “set theoretic expression” is an expression like k>5 ℓ<k Ak+ℓ .

Solution: B = ∞
# "∞
k=1 ℓ=1 (Akℓ ∩ Akℓ2 ).

3
Exercise 5. Let fn , f, g : [0, 1] → [0, 1] and a, b, c, d ∈ [0, 1]. Derive the
following set theoretic expressions:

(a) Show that

{x ∈ [0, 1] | sup fn (x) ≤ a} = {x ∈ [0, 1] | fn (x) ≤ a},

n n

and use this to express {x ∈ [0, 1] | supn fn (x) < a} as a countable combi-
nation (countable unions, countable intersections and complements) of sets
of the form {x ∈ [0, 1] | fn (x) ≤ b}.

(b) Express {x ∈ [0, 1] | f (x) > g(x)} as a countable combination of sets of

the form {x ∈ [0, 1] | f (x) > c} and {x ∈ [0, 1] | g(x) < d}.

(c) Express {x ∈ [0, 1] | lim supn fn (x) ≤ c} as a countable combination of

sets of the form {x ∈ [0, 1] | fn (x) ≤ c}.

(d) Express {x ∈ [0, 1] | limn fn (x) exists} as a countable combination of sets

of the form {x ∈ [0, 1] | fn (x) < c}, {x ∈ [0, 1] | fn (x) > c}, etc. (Hint:
think of {x ∈ [0, 1] | lim supn fn (x) > lim inf n fn (x)}).

Solution: First observe the following set relations

∞
! ∞
[0, c) = [0, c − n1 ] [0, c] = [0, c + n1 )
n=1 n=1
∞
! ∞
(c, 1] = [c + n1 , 1] [c, 1] = (c − n1 , 1].
n=1 n=1

All conversions between strict and non-strict inequalities following from these
relations and properties of the inverse image, i.e. homomorphism of arbitrary
set operations. We will use the shorthand notation

{f < a} := {x ∈ [0, 1] | f (x) < a}.

#
(a) Let x ∈ n {fn ≤ a}. Then, fn (x) ≤ a for all n =⇒ supn fn (x) ≤ a, by
definition of sup as a is an upper bound for {fn (x)}. Therefore, as x was
arbitrary,
∞
{fn ≤ a} ⊂ {sup fn ≤ a}.
n
n=1

4
 Let x ∈ {supn fn ≤ a}. Then supn fn (x) ≤ a and for all n fn (x) ≤
supn (x) ≤ a. Therefore, as x was arbitrary,
∞
{sup fn ≤ a} ⊂ {fn ≤ a}.
n
n=1
#
Hence {supn fn ≤ a} = n {fn ≤ a}. By De Morgan’s this relation also
implies
∞
!
{sup fn > a} = {fn ≥ a}.
n
n=1

Similar results hold for inf.

Let f = supn fn . Using the above comment

{sup fn < a} = {f < a}

n
= f −1 ([0, a))
!∞
= f −1 [0, a − k1 ]
k=1
!∞
= {sup fn ≤ a − k1 }
n
k=1
!∞ ∞
= {fn ≤ a − k1 }.
k=1 n=1

(b) Using countability and density of the rationals

!
{f > g} = {f > q} ∩ {q > g}
q∈Q
!
= {f > q} ∩ {q ≥ g}
q∈Q
!
= {f ≥ q} ∩ {q > g}.
q∈Q

5
(c)

{lim sup fn ≤ c} = { inf sup fk ≤ c}

n→∞ n≥1 k≥n
∞
1
= { inf sup fk < c + m}
n≥1 k≥n
m=1
∞ !∞
1
= {sup fk < c + m}
m=1 n=1 k≥n
∞ ! ∞ !∞
1
= {sup fk ≤ c + m − 1ℓ }
m=1 n=1 ℓ=1 k≥n
∞ ! ∞ ! ∞ ∞
1
= {fk ≤ c + m − 1ℓ }.
m=1 n=1 ℓ=1 k=n

(d)

{ lim fn exists} = {lim inf fn = lim sup fn }

n→∞ n→∞ n→∞
= {lim inf fn < lim sup fn }c (lim inf fn (x) ≤ lim sup fn (x))
n→∞ n→∞ n→∞ n→∞
⎛ ⎞c
!
= ⎝ {lim inf fn < q} ∩ {q < lim sup fn }⎠ (part b)
n→∞ n→∞
q∈Q

= {lim inf fn ≥ q} ∪ {lim sup fn ≤ q}.

n→∞ n→∞
q∈Q

The sets {lim inf n→∞ fn ≥ q} and {lim supn→∞ fn ≤ q} can be expressed
as countable combinations using part (c) and the fact that

− lim sup fn (x) = − inf sup fk (x)

n→∞ n≥1 k≥n

= sup inf (−fk (x))

n≥1 k≥n

= lim inf(−fn (x)),

n→∞

i.e. {lim inf n→∞ fn ≥ q} = {lim supn→∞ (−fn ) ≤ −q}. More specifi-
cally,
" ∞ ∞ ∞ ∞ ! ∞ [ ∞ [
∞ \∞
!#
\ \ [ [ \ \
1 1 1 1
{fk ≥ c − m + ℓ } ∪ {fk ≤ c + m − ℓ} .
q∈Q m=1 n=1 ℓ=1 k=n m=1 n=1 ℓ=1 k=n

6
 Using one of the later two expressions of part (b), we can drop one of the
outer intersections
" ∞ ∞ ∞ ! ∞ [ ∞ [ ∞ \ ∞
!#
\ [ [ \ \
{fk ≥ c + 1ℓ } ∪ {fk ≤ c + m1
− 1ℓ }
q∈Q n=1 ℓ=1 k=n m=1 n=1 ℓ=1 k=n
or
" ∞ [
∞ [
∞ \
∞
! ∞ [
∞ \
∞
!#
\ \ [
1 1 1
{fk ≥ c − m + ℓ} ∪ {fk ≤ c − ℓ} .
q∈Q m=1 n=1 ℓ=1 k=n n=1 ℓ=1 k=n

Exercise 6. Let Ω = N (the positive integers), and let F0 be the collection of

subsets of Ω that either have finite cardinality or their complement has finite
cardinality. For any A ∈ F0 , let P(A) = 0 if A is finite, and P(A) = 1 if AC is
finite.
(a) Show that F0 is a field but not a σ-field.
(b) Show that P is finitely additive on F0 ; that is, if A, B ∈ F0 , and A, B are
disjoint, then P(A ∪ B) = P(A) + P(B).
(c) Show that P is not countably additive on F0 ; that is, construct a sequence of
∞
disjoint sets Ai ∈ F0 such that ∪∞ ∞
i=1 Ai ∈ F0 and P (∪i=1 Ai ) ̸= i=1 P (Ai ).
(d) Construct a decreasing sequence of sets Ai ∈ F0 such that ∩∞
i=1 Ai = Ø for
which limi→∞ P(Ai ) ̸ = 0.
Solution:
(a) The empty set has zero cardinality, and therefore belongs to F0 . Further-
more, if A ∈ F0 , then either A or Ac has finite cardinality. It follows that
either Ac or (Ac )c has finite cardinality, so that Ac ∈ F0 .
Suppose that A, B ∈ F0 . If both A and B are finite, then A ∪ B is also
finite and belongs to F0 . Suppose now that at least one of A or B is infinite.
We have A ∪ B = (Ac ∩ B c )c . Since Ac ∩ B c is finite, it follows that
A ∪ B ∈ F0 . This shows that F0 is a field.
To see
"∞that F0 is not a σ-field, note that {2n} ∈ F0 for every n ∈ N, but the
set n=0 {2n}, the set of even natural numbers, is not in F0 .
(b) Let A, B ∈ F0 be disjoint. If both A and B are finite, then P (A ∪ B) =
0 = P(A) + P(B). Suppose that either A or B (or both) is infinite. Since A
and B are disjoint, we have A ⊂ B c and B ⊂ Ac . It follows that A and B
cannot both be infinite. Therefore, P(A ∪ B) = 1 = P(A) + P(B), and P
is finitely additive.

7
 "
(c) Note that {n} ∈ F0 and n≥1 {n} = Ω. However, P({n}) = 0 while
P(Ω) = 1, hence P is not countably additive.

(d) Let An =#{n, n + 1, . . .}. Then (An )n≥1 forms a decreasing sequence of
sets with n An = Ø. But P(An ) = 1 for all n, hence limn→∞ P(An ) = 1.

8
MIT OpenCourseWare
https://ocw.mit.edu

6.436J / 15.085J Fundamentals of Probability

Fall 2018

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms