MATH50007 Network Science 2021
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
MATH50007
BSc, MSci and MSc EXAMINATIONS (MATHEMATICS)
May-June 2021
Network Science
1. In this question you will work with k-regular Cayley trees with diameter D . Assume that k > 2 and that D > 2. The nodes with degree=1 are termed leaf nodes, and the root node is the node
that is equidistant from each leaf node.
(a) Set k = 3 and D = 4.
(i) Draw the graph (2 marks)
(ii) What is the degree distribution for the graph? (2 marks)
(iii) Assume that x1 , the Katz centrality for the root node, has been computed and is
known. Recall that the Katz centrality for node i in a graph with adjacency matrix A is xi = α Aij xj + 1. Here, α is unknown, but assume that it has been chosen so that the solution is unique. Determine the Katz centralities of the leaf nodes. (3 marks)
(b) Let G0 be a k-regular Cayley tree with k = 3 and D = 4, and let Gt be the graph generated by
the Barabasi-Albert model after t iterations where: 1 link and 1 node are added per iteration, and G0 is the initial graph prior to the first iteration.
(i) What is the probability that a link will be added to one of the leaf nodes in G0 during the first iteration? (2 marks)
(ii) What is the expected number of nodes with degree=3 after 2 iterations? (6 marks)
(c) Consider the following random graph model. Begin with a k-regular Cayley tree with diameter
D . Then, a link is randomly added between each distinct pair of leaf nodes with probability p.
(i) What is the number of links in the graph when p = 1? (2 marks)
(ii) For general p, what is the expected degree of a node that is a leaf node in the initial Cayley tree? (3 marks)
(Total: 20 marks)
2. Consider a complete graph, G, with N nodes. Let A be its adjacency matrix, and let L be its Laplacian. You are given that the eigenvalues of L are 0, N, . . . , N . Consider the ODE
N
x˙i = β (xi _ α) _ α Lij xj , i = 1, 2, . . . , N (1)
j=1
or in vector form
x˙ = β (x _ αl) _ αLx (2)
where l is the N-element vector with all components equal to 1, and α and β are positive constants.
(a) Show that for a complete graph, Ax = Nx _ x and Lx = N (x _ x) where x is the average
of the elements in x . (4 marks)
(b) Show that v1 = l is an eigenvector of L with eigenvalue 0. (1 mark)
(c) Let S be a matrix whose columns are the eigenvectors of L, {v1 , v2 , . . . , vN }, with v1 = l. Let x = Sw, and show that (2) gives
w˙ = Bw + c (3)
where B is a diagonal matrix, and c is a N-element vector whose elements are constants. Clearly describe what the elements of B and c should be. (6 marks)
(d) Given the initial condition, w(t = 0), find the solution, w(t), of equation (3). (4 marks)
(e) Provide a condition for αN which, when satisfied, leads to synchronization of non-trivial
solutions of (1): Ixi (t) _ xj (t)I _ 0 as t _ o. Show that there is synchronization when this condition is satisfied. (5 marks)
(Total: 20 marks)
3. (a) Consider three unweighted, undirected, connected graphs G1 , G2 , and G3 which do not have self-loops or multiedges. Each graph has N nodes and l links, and graph Gi has adjacency matrix Ai . Let G be the graph that consists of G1 , G2 , and G3 collected together (so G is the disjoint union of the three graphs and will have 3N nodes and 3l links). The nodes in G that correspond to G1 are numbered from 1 to N ; nodes corresponding to G2 are numbered from N + 1 to 2N, and the nodes corresponding to G3 are numbered from 2N + 1 to 3N .
(i) What is the adjacency matrix for G? Give your answer in terms of a block matrix which includes the three adjacency matrices A1 , A2 , A3 (3 marks)
(ii) Consider a partition of G into two parts (groups 1 and 2) defined by a vector v o R3N ,
where the ith element of v is 1 if node i is in group 1 and is 0 if node i is in group 2. Provide three mutually orthogonal vectors, {v1 , v2 , v3 } which each correspond to a partition which minimizes the cut size. Each partition should also have at least one node
in each group, and group 2 should have more nodes than group 1.
For any one of the three vectors you have found, show that D1/2v is an eigenvector of
, the normalized adjacency matrix of G : = D- 1/2AD- 1/2 where D is the diagonal
degree matrix for G. Provide the corresponding eigenvalue for this eigenvector. (7 marks)
Consider the following definition of a random walk on a unweighted, undirected, connected
graph, G, with N nodes which does not have self-loops or multiedges. A walker at node i takes a step to node j with transition probability
πij = Aij kj / _ Ail kl _
where Aij is the adjacency matrix for the graph, and kj is the degree of node j .
(i) Provide a concise (1-2 sentence) interpretation of this random walk model. (2 marks)
(ii) Show that this model is equivalent to the standard model for random walks on unweighted
graphs when G is a complete graph. (3 marks)
(iii) Find the stationary probability distribution, po o RN , for this model. Express your
solution in terms of the matrix Mij = ki Aij kj (matrix-vector form is not required). (5 marks)
(Total: 20 marks)
4. Let A be the adjacency matrix for an N-node unweighted, undirected, connected graph with no self-loops or multiedges. The network-SIR model equations for this graph are,
d (si ) = _β N Ail (si xl ) (susceptible)
dt l=1
= µ (xi ) (recovered),
where xi (t) = 1 if node i is infectious at time t and xi (t) = 0 otherwise; si (t) = 1 if node i is susceptible at time t and si (t) = 0 otherwise; and ri (t) = 1 if node i is recovered with immunity at time t and ri (t) = 0 otherwise. The probability that a node which is susceptible at time t becomes infectious at time t + ∆t via a link to an infectious node is β∆t, and the probability that a node which is infectious at time t becomes recovered at time t + ∆t is µ∆t. Note that si (t) + xi (t) + ri (t) = 1.
(a) Show that any infection-free state is an equilibrium state for the system. (2 marks)
(b) Provide a concise (1-3 lines) interpretation of the expression (xi xj ) / (xj ) = 1 in terms of the
states of nodes i and j and how they relate to each other. (2 marks)
(c) Show that the probability that a node is infectious at both times t and t + ∆t is (1 _ µ∆t) (xi (t)) . (4 marks)
(d) Recall that for the network-SI model, the governing equation for the second moment, (xi xj ) is,
d (xi xj ) = β N [Ajl (xi sj xl ) + Ail (si xj xl )] ,
Construct a governing equation for (xi xj ) for the network-SIR model and compare the terms on the right-hand-side of your final equation to those in the 2nd-moment equation for the network-SI model. (6 marks)
(e) Consider the following modification to the network-SIR model above. The probability that
a node that is susceptible at time t is vaccinated and becomes recovered at time t + ∆t is γ∆t. Derive the governing ODEs for (si (t)) and (ri (t)) for this modified model (note that the equation for (xi (t)) will remain unchanged). (6 marks)
(Total: 20 marks)
2022-08-19