Graph Representation Learning.

Catalog description

Semester course; 3 lecture hours (delivered online or face-to-face). 3 credits. Enrollment is restricted to graduate students in the College of Engineering or by permission of the instructor. Students should have prior programming experience and background in linear algebra, probability and basic machine learning. Covers principles and methods for learning on graphs and networks, including problem settings (node/edge/graph prediction), message-passing neural networks, permutation invariance and equivariance, spectral perspectives and positional encodings, and more advanced topics including scalability techniques, graph transformers, and models for heterogeneous and temporal graphs. Provides hands-on experience through assignments and a semester project (empirical, theoretical or systems-focused) emphasizing sound methodology, clear analysis and reproducible artifacts using real-world graph data.