Traitement de données sur graphes (GDP)

• Official homepage (with schedule)

Description

The goal of the graph data processing (GDP) course is to provide a broad introduction to graph algorithms in data science and machine learning.

Content

The course features two modules:

• The network science module. Here we look as the analysis of complex networks from the perspective of network science. We will study the properties of networks and try to understand their underlying mechanics.
• The machine learning on graphs module. A major effort will be given to show that existing data analysis techniques can be extended (and enhanced) on graphs. To do so, we will present mathematical tools based on linear and non-linear graph spectral harmonic analysis.

Keywords

Complex networks, machine learning for graph-structured data, dimensionality reduction, spectral graph theory

Prerequisites

Graph theory, linear algebra, basic probability theory

Acquired skills

At the end of the course, the participants will understand:

• how are complex networks formed
• what are good ways to model them and which of their properties can be predicted
• what is the role of hubs in information spreading and why are epidemics so difficult to contain.

In addition, they will be able to use machine learning techniques in order to process graph-structured data. This entails:

• How can graphs combat the curse of dimensionality?
• Why is the graph spectrum important?
• How to solve problems related to unsupervised, semi- and fully-supervised learning with graphs?

Enseignant

Andreas Loukas