[go: up one dir, main page]

0% found this document useful (0 votes)
189 views3 pages

Notes Compactness

This document provides notes on compact sets in metric spaces. It begins by defining open covers, compactness, and sequential compactness. It then proves that in a metric space, these three notions are equivalent: a set is compact if and only if it is sequentially compact if and only if it is complete and has finite epsilon-covering numbers for all epsilon. It provides examples showing that finite metric spaces and closed bounded subsets of Rn are compact.

Uploaded by

Maja Gwozdz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
189 views3 pages

Notes Compactness

This document provides notes on compact sets in metric spaces. It begins by defining open covers, compactness, and sequential compactness. It then proves that in a metric space, these three notions are equivalent: a set is compact if and only if it is sequentially compact if and only if it is complete and has finite epsilon-covering numbers for all epsilon. It provides examples showing that finite metric spaces and closed bounded subsets of Rn are compact.

Uploaded by

Maja Gwozdz
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Chair for Mathematical Information Science

Thomas Allard
Sternwartstrasse 7
CH-8092 Zürich

Mathematics of Information
Spring semester 2022

Notes on compact sets

Throughout the notes, (X , d) will denote a metric space. These notes are partly based on the
notes [1].

Definitions
Definition 1 (Open cover). Given an index set I and a subset C of X , a family {Oi }i∈I of open
sets Oi ⊆ X is said to be an open cover of C if
[
C⊆ Oi .
i∈I

The family {Oj }j∈J is said to be a finite subcover of {Oi }i∈I if it is a cover of C and contains
a finite number of open sets, i.e., C ⊆ j∈J Oj and J ⊆ I is finite.
S

Definition 2 (Compactness). A subset C of X is said to be compact if every open cover of C


admits a finite subcover.

Definition 3 (Sequential compactness). A subset C of X is said to be sequentially compact if


every sequence {xn }n∈N ⊆ C admits a converging subsequence.

Equivalence of the notions of compactness


We have defined two distinct notions of compactness. It turns out that in a metric space, these
two notions are equivalent and relate to the notion of covering numbers seen in class.

Lemma 1. The following are equivalent:

a) The set C ⊆ X is compact;

b) The set C ⊆ X is sequentially compact;

c) The set C ⊆ X is complete and, for every ε > 0, the ε-covering number of C is finite.

Remark 1. One often encounters the denomination ‘totally bounded’ for a set which, for every
ε > 0, has a finite ε-covering number.

Remark 2. In the lecture, we defined the covering number only for compact sets. The impli-
cation ‘a) =⇒ c)’ in Lemma 1 ensures that the covering number a compact set is well defined
and finite.
Proof. We prove that compactness is equivalent to sequential compactness, which in turn is
equivalent to being complete and totally bounded.
‘a) =⇒ b)’: Assume by contradiction that C is compact but not sequentially compact and
take {xn }n∈N a sequence which admits no converging subsequence. Consider the family

{O ⊆ C | O is open and contains at most a finite number of elements of {xn }n∈N } .

Then this family is an open cover of C. Indeed, for every x ∈ C, since no subsequence of {xn }n∈N
converges to x, there exists an open neighborhood O of x which contains at most finitely many
elements of {xn }n∈N . By compactness of C, it can then be covered by finitely many open sets O,
which each contain at most finitely many elements of {xn }n∈N . We have reached a contradiction.
‘b) =⇒ a)’: Let’s take C to be sequentially compact. Let {Oi }i∈I be an open cover of C.
We first prove that there exists δ > 0 such that, for every x ∈ C, there exists i ∈ I satisfying
B(x, δ) ⊆ Oi . We argue by contradiction, that is, we suppose that we can find a sequence {xn }n∈N
in C such that B(xn , 1/n) is not contained in any Oi . By sequential compactness, {xn }n∈N admits
a subsequence which converges to some x ∈ C. Because {Oi }i∈I covers C, there exists i ∈ I such
that x ∈ Oi , and, since Oi is open, there exists ε > 0 such that B(x, ε) ⊆ Oi . Since {xn }n∈N has
a subsequence converging to x, there exists n ∈ N such that B(xn , 1/n) ⊆ B(x, ε) ⊆ Oi , which
is in contradiction with the construction of {xn }n∈N .
We have proven that there exists δ > 0 such that, for every x ∈ C, there exists i ∈ I satisfying
B(x, δ) ⊆ Oi . Assume, again by way of contradiction, that there is no finite subcover of {Oi }i∈I .
Then, one can construct a sequence {xn }n∈N in C such that xp ∈ / B(xq , δ) for all p 6= q, otherwise
C could be covered by a finite number of balls of radius δ and a fortiori by a finite number
open sets Oi containing these balls (which, we have proven above, exist). The sequence {xn }n∈N
admits a convergent subsequence by sequential compactness of C, but also satisfies d(xp , xq ) ≥ δ
for all p 6= q by construction, which yields a contradiction and concludes the proof.
‘b) =⇒ c)’: Let’s take C to be sequentially compact and {xn }n∈N a Cauchy sequence in C.
By sequential compactness, {xn }n∈N has a converging subsequence. Since {xn }n∈N is Cauchy
and has a converging subsequence, it is convergent in C and therefore C is complete.
Given ε > 0, the family
{B(x, ε) | x ∈ C}
is an open cover of C. We have already proven that sequential compactness implies compactness.
One can thus find a finite subcover to {B(x, ε) | x ∈ C}. The cardinality of a minimal such
subcover is, by definition, the ε-covering number of C, which is therefore finite.
‘c) =⇒ b)’: Let’s assume that C is complete and has finite (1/m)-covering number for every
integer m ≥ 1. We consider a sequence {xn }n∈N in C and want to show that it admits a
(1) (1)
converging subsequence. Let {y1 , . . . , yN1 } be a 1-covering of C, then, by finiteness of the
(1)
covering, one can extract a subsequence {xn }n∈N of {xn }n∈N such that all its elements are in
(1) (m) (m−1)
B(y` , 1) for some 1 ≤ ` ≤ N1 . Likewise, one can find a subsequence {xn }n∈N of {xn }n∈N
(m) (m) (m)
such that all its elements are contained in B(y` , 1/m), where {y1 , . . . , yNm } is a (1/m)-
(k)
covering of C and 1 ≤ ` ≤ Nm . The sequence {xk }k∈N is therefore a Cauchy sequence. By
completeness of C, it is a converging subsequence of {xn }n∈N , which completes the proof.

Examples
Corollary 1 (Finite metric space). A metric space (X , d) containing a finite number of elements
is compact.
Proof. Since X is finite, for any sequence {xn }n∈N ⊆ X there must exists a point x ∈ X such
that {xn }n∈N passes infinitely many times through x. Writing {nk }k∈N the indices for which
xnk = x, we have constructed a subsequence {xnk }k∈N which converges (to x). Therefore X is
sequentially compact. From the implication ‘b) =⇒ a)’ in Lemma 1, the set X is compact.

Example 1. Given n ∈ N, let the Hamming cube be defined as

Hn := {0, 1}n ,

and let

d : Hn × Hn → R+
(x, y) 7→ #{i ∈ {1, . . . , n} | xi 6= yi }

be a metric on Hn . Then Hn is finite and therefore compact. From the implication ‘a) =⇒ c)’
in Lemma 1, the set Hn has a finite covering number. We refer the reader to [Problem 4, Exam
Winter 2021/2022] for an estimation of this covering number.

Corollary 2 (Closed bounded sets in Rn ). A closed bounded subset C of Rn equipped with the
Euclidean norm k · k2 is compact.

Remark 3. Since Rn is a finite dimensional vector space, all the norms on Rn are equivalent.
Therefore, the result above generalizes readily to Rn equipped with any norm k · k.

Proof. Consider a Cauchy sequence {xn }n∈N ⊆ C. It converges to some x ∈ Rn by completeness


of Rn . Since C is closed, x ∈ C which proves that C is complete.
Fix ε > 0. Since C is bounded, it is included in the cube [−R, R]n . From the lecture, we
know that the ε-covering number of this cube is finite. To show that it implies that C is totally
bounded, we make the following observation. Since the ε-covering number of the cube is finite,
there exist x1 , . . . , xN ∈ Rn such that [−R, R]n ⊆ ∪Ni=1 B(xi , ε/2). Then, for every i ∈ {1, . . . , N }
such that C ∪ B(xi , ε/2) 6= ∅, choose ci ∈ C ∪ B(xi , ε/2). It follows from B(xi , ε/2) ⊂ B(ci , ε) that
C ⊆ ∪i B(ci , ε). The set C is thus totally bounded.
C is complete and totally bounded, therefore, from the implication ‘c) =⇒ a)’ in Lemma 1,
C is compact.

References
[1] Giada Franz. Equivalent Notions of Compactness.

You might also like