Federica Milinanni: Pre-colloquium

Machine learning aims at building models based on data, which in numerous fields of applications are represented by graphs. To be able to develop statistical models involving graphs it is, therefore, crucial to understand their structure and representation. In particular, it is useful to work with equivalence classes of graphs under the relation of permutations of nodes.

In this talk we will discuss about graphs and their representation via adjacency matrices. We will introduce group actions, that will be used to define the above-mentioned equivalence classes of graphs and the related quotient space, known as graph space. Next, we will introduce some statistical tools to analyse data, such as principal component analysis (PCA), regression models, and deep leaning methods. 

This is a pre-colloquium targeted towards master and PhD students as a preparation for Aasa Feragen's colloquium talk  at 15:15.