The seminar concerns graph neural networks and their applications. Graph neural networks are an increasingly popular type of deep neural networks in which the input data comes from a certain graph (they are observations of certain fragments of a given graph), the network is a model of the given graph, and the learning algorithm constructs this model based on training data (recorded observations). The aim of the seminar is to discuss the basic types of graph neural networks and their learning algorithms, including networks for classifying graph nodes, classifying entire graphs, networks for predicting connections between nodes in a graph, predicting the values of graph node features, etc., as well as their contemporary applications in practice.
