Conference: Quantitative Network Science


Learning over networks for data-driven science

{{_Ltalk:R}} Dr. Dorina Thanou
Date: 26.11.20   Time: 09.00 - 09.45   Room:

The past decade has seen a significant amount of interest in utilising graphs, mathematical tools for capturing pairwise relations (i.e., edges) between entities (i.e., nodes), to represent complex network structures. Going beyond only information contained in the edges, recent years have seen increasing attention paid to the representation and processing of data collected in network domains, i.e., data observed on the nodes of the graph.

In this talk, we focus on some recent advances in modeling and analysing network data. We first provide ways of incorporating the network structure into typical representation learning algorithms such as node embeddings and dictionary learning. We then focus on the very important problem of inferring hidden relationships from data and provide a unified framework inspired by the emerging field of graph signal processing. Finally, we provide applications of the above mentioned algorithms in different domains such as computer vision and healthcare.