Selected Topics on Deep Learning over Structured Data

This course studies machine learning methods for structured data, with an emphasis on relational databases, tables, and graphs. We will examine how these data representations are connected, and how learning methods can move between database-specific, schema-aware, and cross-schema settings.

Topics include feature engineering and automated feature generation, vector embeddings, graph neural networks, transformer-based foundation models, and explainability methods for model predictions. Tutorial sessions will focus on implementing the covered methods using modern libraries for machine learning, deep learning, and graph processing.