Scribd
Upload
11 views2 pages

MMD Hw2

Mmd

Uploaded by

manhn031103
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
11 views2 pages

MMD Hw2

Mmd

Uploaded by

manhn031103
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/ 2

Homework 2

Instructions:

• Due Date: January 08th, 2024

• Submission:

– Submit your solutions as a printed document.


– Submit directly to the lecturer during the final exam.

Grading Criteria:

• Clarity and Completeness: Clear explanation of steps and reasoning.

• Accuracy: Correct calculations and results.

• Insight: Demonstration of understanding of underlying concepts through discussion


and analysis.

1. PageRank
Consider a directed graph with 5 nodes:

• Node A links to Nodes B and C.

• Node B links to Nodes C and D.

• Node C links to Node A.

• Node D links to Node C.

• Node E links to Nodes B, D, and C.

Using the PageRank algorithm with a damping factor of 0.85, calculate the PageRank values
for each node after 3 iterations. Show your calculations step by step.

2. Node Embedding
Describe the role of node embeddings in graph analysis. Using the Karate Club dataset as
an example, explain how Node2Vec generates embeddings for nodes. Illustrate your expla-
nation with a simple example of a walk generation process and the role of hyperparameters
like p and q.

1
3. Graph Convolutional Network (GCN)
Given a simple undirected graph with 4 nodes and the following adjacency matrix:
 
0 1 1 0
1 0 1 1
A= 1 1 0

1
0 1 1 0

Assume each node has an initial feature vector X of dimension 2:


 
1 0
0 1
X= 1 1

0 0

Calculate the output features for each node after applying one layer of a GCN with a
weight matrix:  
1 1
W =
1 1
Normalize the adjacency matrix using  = D−1/2 (A + I)D−1/2 , where D is the degree
matrix, and show all intermediate steps.

4. Optimizing Submodular Functions


Suppose you are tasked with selecting k influential users from a social network to maximize
the spread of information using a greedy algorithm. Explain why the problem can be formu-
lated as optimizing a submodular function. Use the following hypothetical influence scores
for subsets of users:

• f ({A}) = 3

• f ({B}) = 5

• f ({A, B}) = 7

• f ({A, C}) = 6

• f ({B, C}) = 8

Run the first two steps of the greedy algorithm to select k = 2 users and justify your
choices based on marginal gains.

You might also like