Node Vector Distances: Methods and Applications

Michele is an associate professor at IT University of Copenhagen since 2020. He mainly works on algorithms for the analysis of complex networks, with methods spanning from Computer Science to physics, and on applying the extracted knowledge to a variety of problems, including economic development, human mobility, and the analysis of ideological polarization. His background is in Digital Humanities, connecting new technologies with a classical education. He has a PhD in CS, obtained in 2012 at the University of Pisa. In the past, he visited Barabasi’s CCNR at Northeastern University, and worked with Ricardo Hausmann for 6 years at the Center for International Development, Harvard University.

–

How quickly is a disease spreading through a social network? How diversified is a country in a network of related exported products? How polarized is the online discourse on social media? These seemingly unrelated questions can be answered with a common toolbox: they all require to compare different types of node attributes. In this workshop, you will learn how to apply Node Vector Distance techniques: given two node attributes — represented as vectors — how far the two are on a network structure? Are they correlated when we take into account not only the value of a node itself, but also the values of its neighbors? Or simply: how spread out in the network is an attribute? The workshop will make you familiar with ways to translate concepts such as Euclidean distance, Pearson correlation, and variance, when the vectors do not live in an Euclidean space, but on the complex landscape that is the topology of a network.