Title:

On statistical learning of graphs

Abstract:

In the first part of this talk, I aim to give an introduction to two frameworks that describe when a family of functions can be considered to be learnable, namely PAC and online learnability, and discuss the relationship between them. In the second part, we will apply these frameworks to study the graphs such that the family of their isomorphic copies (with some constraints to be made precise) are learnable.

This is joint work with Vittorio Cipriani, Valentino Delle Rose, and Luca San Mauro.