You are on page 1of 2

Random Graphs

Assignment - 9

May 11, 2020

1. Recall that
(
1+δ
(1,δ) t(2+δ)+(1+δ) for i = t + 1
P (vt+1 → vi |Gt ) = di (t)+δ (1)
t(2+δ)+(1+δ) for i ∈ [t]

Show that the attachment probabilities defined above sum up to one.


(t)+δ Qt−1
2. Define Mi (t) = di1+δ (2+δ)k+1+δ
k=i−1 (2+δ)(k+1) . Show that {Mi (t)}t≥i is a
non-negative Martingale with mean 1.
3. Define:
(
0 for k = 0, . . . , m − 1
pk = Γ(k+δ)Γ(m+2+δ+δ/m)
(2 + δ/m) Γ(m+δ)Γ(k+3+δ+δ/m) for k ≥ m

Show that {pk }k≥0 is a probability mass function.

4. Complete the proof of the main theorem (given below) using propositions
4.3 and 4.4 stated in notes.

Theorem 0.1 Fix m ≥ 1 and δ ≥ −m. There exists a constant C =


C(m, δ) > 0 such that as t → ∞,
r !
log t
P max |Pk (t) − pk | ≥ C = o(1)
k t

5. Show that
p 2 2
P (|Nk (t) − E[Nk (t)]| ≥ C t log t) ≤ 2e−(log t)C /(8m )

6. Consider a preferential attachment model P A0t (1, δ) without self-loops de-


fined as follows:

1
• P A01 (1, δ) consists of a single vertex v1
• P A02 (1, δ) consists of two vertices v1 and v2 with connected by two
edges between them.
• At time t > 2 for all i ∈ [t]

di (t) + δ
P (vt+1 → vi |P A0t (1, δ)) =
t(2 + δ)

Show that the law of {P A0t (1, α)}t≥1 is equal to one of {P At (1, δ)}t≥2 for
δ
α = 2+δ . (Hint: Use induction on t ≥ 1.)
7. Write a recursion for Nk (t), number of vertices of degree k at time t for
the above preferential attachment model P A0t (1, δ).
8. Let Nt denote the number of connected components of P At (1, δ). Show
that for δ > −1,
 
1+δ 1+δ
E[Nt ] ≤ 1 + log t +
2+δ 2+δ

You might also like