Scribd
Upload
2 views3 pages

C8.4_problems3

This document contains exercises related to probabilistic combinatorics, focusing on topics such as Chernoff bounds, branching processes, and tournaments. It includes problems on the distribution of edges in bipartite graphs, discrepancies in hypergraphs, and survival probabilities in Poisson processes. Additionally, it discusses the relationship between extinction probabilities and branching processes, as well as optional exercises for further exploration.

Uploaded by

Richard Zhang
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
2 views3 pages

C8.4_problems3

This document contains exercises related to probabilistic combinatorics, focusing on topics such as Chernoff bounds, branching processes, and tournaments. It includes problems on the distribution of edges in bipartite graphs, discrepancies in hypergraphs, and survival probabilities in Poisson processes. Additionally, it discusses the relationship between extinction probabilities and branching processes, as well as optional exercises for further exploration.

Uploaded by

Richard Zhang
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

C8.

4 Probabilistic Combinatorics
Sheet 3 — HT24
Chernoff bounds; Branching processes

Section B
1. If (V1 , V2 ) is a fixed partition of the vertices of G(n, 1/2), what is the distribution of the
number of edges of G(n, 1/2) joining V1 to V2 ?
Show that the probability that G(n, 1/2) contains a bipartite subgraph with at least
n2 /8 + n3/2 edges is o(1).

2. A tournament on a vertex set V is an orientation of the edges of the complete graph


on V . Thus, for each pair {i, j} of distinct elements of V , exactly one of the directed
edges i → j and j → i is present. (Think of an all-play-all tournament whose players
are the elements of V ; the orientation of the edge between i and j indicates who wins
the match between i and j.)
Let σ be a permutation of {1, 2, . . . , n}. The permutation can be seen as a ranking of
the players. We say that an upset occurs if the edge i → j is present (i.e. j beats i) but
σ(i) < σ(j) (i.e. i is higher ranked than j).
Show that there exists a tournament on {1, 2, . . . , n} such that, for all rankings σ, the
difference between the number of upsets and the number of non-upsets is no greater than
√
2n3/2 log n. (In other words, no ranking gives a correct prediction for significantly more
than 50% of the matches.)

3. Let H = (V, E) be a hypergraph. Let χ be a two-colouring (red/blue) of its vertices.


The discrepancy of an edge e ∈ E under the colouring χ is the absolute difference
between the number of blue vertices in e and the number of red vertices in e. The
discrepancy of H under χ, denoted disc(H, χ), is the maximum over all edges e of
the discrepancy of e under χ. Finally, the discrepancy of H, disc(H), is defined as
minχ disc(H, χ).
[For example, if H is k-uniform, disc(H) < k if and only if H is 2-colourable.]
√
(a) Show that if H is k-uniform and has m ⩾ 2 edges, then disc(H) ⩽ 2 k log m.
(b) Show that if H is k-uniform and each edge intersects at most d other edges, then
p
disc(H) ⩽ 2k log(6(d + 1)).

Mathematical Institute, University of Oxford Page 1 of 3


Oliver Riordan: [email protected]
C8.4 Probabilistic Combinatorics: Sheet 3 — HT24

4. Using results from lectures, show that the survival probability ρ(c) = 1 − η(c) of the
Poisson branching process XPo(c) satisfies ρ(1 + ε) ∼ 2ε as ε tends to zero from above.
Can you obtain further terms in this expansion?

5. Let Yk denote the number of k-vertex components of G = G(n, p).


(a) Using Cayley’s formula k k−2 for the number of trees on k (labelled) vertices, show
directly that if k is fixed and p = p(n) satisfies np → c with c > 0 constant, then
 
n k−2 k−1 −ck
EYk ∼ k p e .
k

(b) Deduce that ρk (c) = ck−1 k k−1 e−ck /k!. [You may like to give a direct proof of this
formula.]
(c) Deduce that
∞
X k k−1 −ck
ck−1 e =1
k=1
k!
if 0 ⩽ c ⩽ 1, and that the sum is strictly less than 1 if c > 1. [You may not like to
give a direct proof of this!]

6. (a) Show that for each c ∈ (1, ∞) there is a unique d ∈ (0, 1) such that ce−c = de−d .
(b) Let η be the extinction probability of XPo(c) , the Galton–Watson branching process
with offspring distribution Po(c). Show that cη = d where d is related to c as in
part (a).
(c) Consider the first particle (the root) in the branching process XPo(c) . What is the
probability of extinction of the process conditional on the event that the root has
k children (for k ∈ {0, 1, 2, . . . })? Use this to find the conditional distribution of
the number of children of the root, conditional on the event that the process dies
out.
(d) Hence or otherwise argue that the branching process XPo(c) , conditioned on extinc-
tion, has the same distribution as the branching process XPo(d) .
What does this suggest about the random graphs G(n, d/n) and G(n, c/n)?

Mathematical Institute, University of Oxford Page 2 of 3


Oliver Riordan: [email protected]
C8.4 Probabilistic Combinatorics: Sheet 3 — HT24

Section C
These questions are optional, but MFoCS students are recommended to at-
tempt them. Outline solutions will be provided later; you can also approach
class tutors (if they have time) or the lecturer to discuss them.

7. (a) Let X1 , X2 , . . . , Xn be independent random variables such that 0 ⩽ Xi ⩽ 1. Let


Sn = ni=1 Xi and let p = EXi /n, so that ESn = np. Show that
P P

P (Sn ⩾ xn) ⩽ e−uxn (1 − p + peu )n

for any u > 0, x > p, and deduce that the Chernoff bounds proved in lectures for
the case Sn ∼ Bin(n, p) also apply in this more general case.
(b) Let a1 , . . . , an be constants and let c > 0. Let Y1 , . . . , Yn be independent random
variables such that ai ⩽ Yi ⩽ ai + c, for all i. Give (with brief justification) a
version of the Chernoff bound for P(Sn − ESn ⩾ t), where Sn = ni=1 Yi .
P

Mathematical Institute, University of Oxford Page 3 of 3


Oliver Riordan: [email protected]

You might also like