CSL7870: Learning on Graphs and its Applications

Offered in Fall 2025 Semester

Credits

The course is a 3 Credit course with L-T-P : 3-0-0 and offered at 700 Level. That is 3 contact hrs and 9 non-contract hrs a week.

Slot N

The course will run in N-slot. The timings are Tuesday, Thursday 8:00AM - 8:50AM and Friday 5:00PM - 5:50PM

LMS

Moodle Link: SoNAA Learning Hub Joining code is: Ruined-Proofing. Login with google, use the institute email.

Syllabus

Introduction: Complex Systems, Graph Data Structure, Graph Representations, Spectral graph theory, Real world networks and applications.

Graph Representation Learning: Node Embeddings, Graph Convolutional Networks, Design Space, Inductive Representation Learning, Graph Attention Networks, Hierarchical Graph Representation Learning, Graph Embeddings, Over smoothing Issues, Expressiveness of GNN, GNN vs CNN, Graph Representation Learning vs Graph Signal Processing.

GNN Applications: Heterogeneous and Multilayer graphs, Knowledge graphs, Reasoning over knowledge graphs, Graphs and LLMs, Neural subgraph matching, GCN for Recommender Systems, Node Classification, Graph Classification, Community Detection, Link Prediction and Causality.

Large Scale Graphs: Dealing with large graphs, Cluster-GCN, Simplifying Graph Convolutional Networks.

Learning Materials

Textbook

Reference Books

  • Network Science, by Albert-Laszlo Barabasi, (Cambridge University Press - August 2016) freely available under the Creative Commons licence. http://www.networksciencebook.com/

Self Learning Material

Grading Policy
Mini Tasks Team Learning Mini Project Minor + Major
25% 10% 10% 15% + 40%

Mini Tasks

Mini Tasks are a central hands-on component of the course. Through these tasks, students will learn the basics of the subject by engaging in various activities such as program implementation, data set preparation, and visualization. For the first 8 to 9 weeks, students will work on these mini tasks. One mini task will be available corresponding to each lecture, and the students need to submit it within a short span (say within 24 hrs). The expected time will vary from 30 min to 4 hrs, and depending on the same, the deadlines will be decided. I will try to make sure that the out-of-class engagement remains less than that of the requirement of a 3Cr, 700-level course (i.e., 9 hrs a week).

Mini Project

This will be a large-scale (as compared to mini tasks) project where students will work in groups to achieve some goal. This project will run during the 8th-12th week of the course. There will be a demo/presentation and viva for evaluation in the last week of the course. It is mandatory for each group to meet with me/TA for discussion on the progress each week. Missing the meeting (by a group or individual) will negatively affect the grading (of the group or individual).

Team Learning

To foster collaborative learning, students will be randomly assigned to a team of about five members. Size will be finalized after the add-drop date. Mean pooling will be used as the score for the team. Students are encouraged to help team members learn the concepts and score well on tasks and projects. The mean-pooled score will be allocated to each team member as a grading component.

Minor and Major

Minor exam will be with pen and paper, offline, closed book. The questions will be mixed of objective and subjective. Logistics of the major will be declear after minor. It will be either same as Minor or Open Book Problem Solving based.

Gems and Leaderboards
On completion of each activity (tasks, polls, attendance, or many others in the LMS), a student will be awarded Gems (Graph Embedding Molecules) points. A student will climb levels with Gems. The LMS will have two leaderboards: one for the individual leaderboard and another for teams. A detailed rule will be available in the LMS and Zulip.
Attendance Policy
As per the institute policy. You are expected to attend 75% of classes. Maximum two weeks of continuous leave. In other words I expect you to see in the classroom for about 28 to 30 classes out of 39. Those taking long leave, for more than 5 classes, please inform me or course TA.
Plagiarism tolerance

Plagiarism tolerance is 7% from single source and 15% cumulative, anything more will reduce your marks as follows:

  • Any logical/conceptual/formulation plagiarism: zero marks
  • Other form of plagiarism (above 50%): zero marks
  • Otherwise: Percentage of plagiarism will be deducted from the obtained mark

These policy will apply for any assignments, hands on and practical work. Copying from friends and internet both treated similarly.