Machine Learning for Complex Networks

Graph representations of relational data have become an important foundation to address data science and machine learning tasks across the sciences. Graph mining and learning techniques help us to detect functional modules in biological networks and communities in social networks, to find missing links in social networks, or to address node-, link-, or graph-level classification tasks.

This course equips students with techniques to address supervised and unsupervised learning tasks in data on complex networks. We show how statistical learning techniques can be used to infer cluster patterns or predict links, introduce methods to learn low-dimensional vector-space representations of graph-structured data, and discuss applications of deep learning to complex networks.

The course combines a series of lectures -- which introduce theoretical concepts in statistical learning, representation learning, or graph neural networks -- with practice sessions that show how we can apply them in practical graph learning tasks.

The course material consists of annotated slides for lectures and a series of accompanying jupyter notebooks. Students can apply and deepen their knowledge through weekly exercise sheets. The successful completion of the course requires to pass a final written exam.