Week 1 Notes
Week 1 Notes
Week 1 Notes
We will start with a simple combinatorial problem. Consider {−1, 1}1000 . How many
elements
x ∈ {−1, 1}1000
satisfy
¯X ¯
¯ 1000 ¯
¯ xi ¯¯ ≥ 50?
¯
i=1
x ∈ {−1, 1}n
satisfy ¯X ¯
¯ n ¯
¯ x ¯ ≥ λn?
¯ i ¯
i=1
The answer is given by the binomial distribution. We are only seeking approximations.
This is a question that we will spend a fair deal of time on this quarter. Today we
will be satisfied with a crude upper bound.
Fact: For any r ∈ R
Pn
For x ∈ {−1, 1}n we write Sn (x) i=1 xi and for m < n we write x|m for the restriction
of x to the first m terms.
Lemma 1.0.1 X
(Sn (x))2 = n2n .
x∈{−1,1}n
1-1
Lecture 1: Introduction 1-2
For each y ∈ {−1, 1}n+1 we will write it as a pair (x, a) where x ∈ {−1, 1}n and
a ∈ {−1, 1}.
à n+1 !2
X X X
yi = (Sn (y|n ) + yn+1 )2
y∈{−1,1}n+1 i=1 y∈{−1,1}n+1
X
= (Sn (X) + 1)2 + (Sn (x) − 1)2
X∈{−1,1}n
X
= 2 (Sn (x))2 + 1
x∈{−1,1}n
= 2(n2n + 2n )
= (n + 1)2n+1 .
Then X X
n2n = Sn (x)2 ≥ s(x)2 ≥ |Aλ,n |(λn)2 .
x∈{}n x∈Aλ,n
1. variance
2. large deviations
3. Chebychev’s inequality
Lecture 1: Introduction 1-3
There is one real drawback of the approach that we took. It works fine for finite
statements but it doesn’t allow us to make statements about infinite objects like
{−1, 1}∞ .
What questions do we want to ask? For example what fraction of x ∈ {−1, 1}∞
satisfy
n
1X
lim xi → 0?
n→∞ n
i=1
Since the set of x which satisfy this condition is infinite, as is the set of x which
don’t satisfy the condition. We cannot answer this question like we did with counting
arguments above. With measure theory this question will be easy to state.
But now we show how to give an answer to a similar question which can be analyzed
by combinatorial means.
Define
½ ¾
N 1
B²,N0 ,N = x ∈ {−1, 1} : there exists m such that n0 < m ≤ N with Sm (x)| > ² .
m
We wish to show that for every ² > 0 there exists N0 such that 21N |B²,N0 ,N for all
N > N0 . This is a finite version that indicates that for “most” x ∈ {−1, 1}N we
have that n1 Sn (x) → 0. We won’t quite be able to do this today but there is a slight
modification of our argument can be made to work.
Fix ² > 0, N0 and N . For every x ∈ B = B²,N0 ,N there exists an m such that x|m is
in A²,m . And for every element of A²,N there are exactly 2N −m elements of B. Thus
by our previous estimate
N
X N
X XN
N −m 2m N −m 2N
|B| ≤ |A²,m |2 ≤ 2 ≤ .
m=N0 +1 m=N0 +1
²n m=N +1
²m
0
PC
If the series m
were summable then we could just choose N0 large enough so that
the sum were less than ² and be done. Unfortunately that isn’t true. In about two
weeks we will return to this argument to show how it modify this argument to make
it work.
Lecture 1: Introduction 1-4
We will define
½
1
B̃²,N0 ,N = x ∈ {−1, 1}N : there exists m such that n0 < m ≤ N with Sm (x)| > ²/2
m
¾
and m is a perfect square .
1
1. for every ² > 0 there exists N0 such that 2N
|B̃²,N0 ,N | < ² for all N > N0 and
2. B²,N0 ,N ⊂ B̃²,N0 ,N .
Example 2.1.3 The position of a body in a 3-D Euclidean space belongs to the set
Ω = R3 .
1. Ω ∈ F
2. A ∈ F implies Ac ∈ F
3. A, B ∈ F implies A ∪ B ∈ F
Recall that A ∩ B = (A ∪ B)c . Thus, a set of subsets F that is closed under comple-
ments is is closed under unions if and only if it is also is closed under intersections.
Therefore, closure under union in the definition above could be replaced by closure
under intersection.
2-1
Lecture 2: Ideas from measure theory 2-2
1. F is a field.
2. A1 , A2 , ... ∈ F implies A1 ∪ A2 ∪ .... ∈ F.
A field is closed under finite set theoretic operations whereas a σ-field is closed un-
der countable set theoretic operations. In a problem dealing with probabilities, one
usually deals with a small class of subsets A, for example the class of subintervals
of (0, 1]. It is possible that if we perform countable operations on such a class A of
sets, we might end up operating on sets outside the class A. Hence, we would like to
define a class denoted by σ(A) in which we can safely perform countable set-theoretic
operations. This class σ(A) is called the σ-field generated by A, and it is defined as
the intersection of all the σ-fields containing A (exercise: show that this is a σ-field).
σ(A) is the smallest σ-field containing A.
1. 0 ≤ P(A) ≤ 1 for A ∈ F.
2. P(∅) = 0, P(Ω) = 1.
S P
3. If {Ai }i∈N are disjoint events in F, then P( i Ai ) = ∞
i=0 PAi .
Example 2.1.8 If A is the class of subintervals of Ω = (0, 1), then the sigma field
generated by A, denoted by B, is called the collection of Borel sets of the unit interval.
The probability space on a unit interval is then defined as (Ω, B, P), where Ω = (0, 1),
P(B) = λ(B) for B ∈ B. Here λ denotes Lebesgue measure, for which λ((a, b]) = b−a.
A = {x : x1 = a1 , . . . , xn = an }
or
µ(A) = p#{i:ai =1and i is prime} (1−p)#{i:ai =0}and i is prime q #{i:ai =1and i is not prime} (1−q)#{i:ai =0}and i is not prime