Universiteit Leiden

nl en

Lecture

LCN2 seminar November 2024

Date
Friday 29 November 2024
Time
Address
Gorlaeus
Room
BM. 2.26 (second floor)

70th LCN2 seminar

Speaker: Riccardo Milocco (IMT Lucca) 

Title: Multi-Scale Node Embeddings for Graph Modeling and Generation

Abstract: 
Networks have successfully represented a variety of complex systems, from economic interactions to ecosystem activities. Yet, in general, they do not provide a natural "embedding space'', from which a set of node coordinates can be extracted. 

Since machine-learning applications, e.g. link prediction, require node features as inputs, node-embedding methods have become increasingly important to generate these attributes. However, typical embedding schemes regard the network only  at the scale where the data were tracked down. If nodes were merged into a community, then the embedding for the community would have to be recomputed; there is no way to obtain it from the embeddings of its members. To address this limitation, we enhanced the Multi-Scale Model by equipping it with node embeddings. We tested our new multi-scale embedding model against a single-scale model using two multi-scale datasets: the ING Bank Input-Output Network and the World Trade Web. The results demonstrate that single-scale models are affected by "single-scale overfitting'', providing descriptions of economic phenomena that are not generalizable at coarser scales. In contrast, the multi-scale model remains self-consistent at every lower-resolution level, hereby offering a more comprehensive and robust representation of the economic system. Although we will focus on the problem of network modeling, the problem of robust embeddings is also crucial for many other applications such as link prediction, community detection, and network visualization.

Snacks and drinks will be available after the seminar at the Fusiebar.

You can also attend online

Zoom link
This website uses cookies.  More information.