Exercises and Assignments
Pen and Paper: Invent the Definitions Before You Are Given Them
A pencil, two five-student friend groups, and no algorithm. The first half asks you to say exactly how much imperfection a “group” may contain — and in the process you reconstruct the k-plex, the k-core, the \rho-dense subgraph, the n-clique and the k-truss yourself. The second half hands you eight students in a ladder and asks you to cut it in two: first the cheapest cut, then the cheapest balanced cut, which is where the ratio cut and the normalized cut come from.
- Discovering Friend Groups (handed out in class; the PDF is being rewritten)
You work this in class, before any of it is named. Everything the session goes on to formalize, you will have already argued about with your own hands.
Be the Algorithm
The same tool as the concepts page, on all four networks. In each one you click a node to repaint it, the score updates on the spot, and the dashed line is the best score known for that network with that many colours.
- Two cliques, modularity — the warm-up. Reach the dashed line.
- Two cliques, graph cut — now you are minimizing the number of edges cut. Beat the dashed line, and then look at what you built.
- Karate club — a real network, four colours.
- Two cliques plus a big one — the resolution limit — three colours for what are obviously three groups. Try the obvious answer first, then find the one that scores higher.
- A random network with no communities at all
Two things to take from the list. On the resolution-limit network, the partition that maximizes Q is not the one your eye picks: it merges the two small cliques, and the dashed line is set where the merge is. On the random network, the best Q you can reach is higher than the best Q on the two honest cliques. If a random network outscores a real one, what does a modularity score on its own actually tell you?
Assignments
Pair Notebook
Work through this module with the AI tutor and hand in the notebook that the session builds. What is graded is your reasoning, not your code — see the Pair Notebook guide for the setup and a walkthrough of a full session.
- Repository for M05: coming soon. Your invitation link arrives through Classroom 50.
Group Mini-Project
Groups of up to three. One repository per group, graded once, with the same score going to every member of the group. Read how a mini-project runs first — exactly one person accepts the assignment and invites the rest, and that is the step teams get wrong.
- Repository for M05:
adv-net-sci-mini-proj-clustering. Accept the invitation through Classroom 50 first — the repository is not public, so the link only opens once you are enrolled.
Readers who are not enrolled in SSIE 641 cannot open the assignment repositories. The worksheet above and the hands-on notebook cover the same material and need no enrollment.