Graph Deep Learning
The goal of this course is to provide the fundamentals of graph deep learning by presenting the core components and related applications.
Graphs are structures commonly found across science to describe the complex relations that exist among the entities of a system, like chemical bonds between atoms, friendships and personal relations on social media, the spread of diseases among countries. Recent advances in deep learning have focused on developing neural networks able to process graph-based structures by leveraging on node and relational information to solve a machine learning task. Students will learn how to process graphs by embedding them to vector spaces for traditional and deep processing as well as design and implement graph convolutional networks. The course will also cover some basic theoretical aspects to understand the underlying mechanisms.
In addition to frontal lectures, practical laboratory sessions in Python will provide the the due hands on.
Assignments and exam.
- Deep Learning Lab
- Machine Learning
- Material provided by the instructors
- D. A. Spielman - Spectral and Algebraic Graph Theory (http://cs-www.cs.yale.edu/homes/spielman/sagt/) • Bruna et al. - Spectral Networks and Deep Locally Connected Networks on Graphs (https://arxiv.org/abs/1312.6203)
- Defferrard et al. - Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering (https://arxiv.org/abs/1606.09375)
- Kipf & Welling - Semi-Supervised Classification with Graph Convolutional Networks (https://arxiv.org/abs/1609.02907)
- Gilmer et al. - Neural Message Passing for Quantum Chemistry (https://arxiv.org/abs/1704.01212)
- Xu et al. - How Powerful are Graph Neural Networks? (https://arxiv.org/abs/1810.00826)