I work on two tighly connected topics. How to define and quantify resilience, in complex systems in general and in social organisations in particular. How to model complex systems from the data, to perform a data-driven study of their properties.
Resilience of Social Organisations
We aim at an understanding of how different online social organisations withstand and recover from internal and external shocks, and how such a resilience can be improved. To do, so we investigate how different resilience metrics can be computed from data. Our research is deeply rooted in data science and the data-driven modelling of complex systems. We use and develop new methodologies from network science, multivariate statistics, machine learning, and statistical physics.
Statistical Models for Network Data
To allow the study of complex dynamical systems such as social organisations, we need inferential models aimed at temporal and multi-edge network data. I am interested in the development of quantitative methods for the analysis of repeated interactions arising in social, economical, and political networks. A second line of research I follow focuses on modelling and understanding temporal correlation observed in temporal network data, such as distribution networks and communication networks.
Part of my work revolves around the generalised hypergeometric ensemble of random graphs, gHypEG for short. In its simplest form, gHypEG provides a model preserving vertices' activities. Doing so, it maps the standard configuration model to an urn problem. Pairs of nodes are like balls in an urn. The more frequent are specific balls, the more likely are edges to be sampled. In its general form, edge probabilities are again defined by balls' frequencies, but also by independent edge propensities estimated from data. The higher the propensity, the larger the ball, the easier it is to sample the edge. The main applications of gHypEG, are towards the inference of significant relations from observed interactions, and the analysis of complex networks by means of network regressions.
The R package ghypernet provides an Open Source implementation for R of a set of functions to work with gHypEG models.
GHYPERNET Tutorials and Material
The following links contain a collection of tutorials and material about the ghypernet R package.
Tutorial to network regression models: https://sg.ethz.ch/nrm-tutorial/
Companion git repository for the 2019 EUSN Workshop in Zurich: https://github.com/sg-dev/EUSN2019_r-ghypernet
Repository for the EuroCSS Tutorial “Introduction to Multi-edge Network Inference in R Using the Ghypernet-package”: https://github.com/sg-dev/EuroCSS2019_r-ghypernet
GitHub repository for ghypernet: https://github.com/gi0na/r-ghypernet
Manual and Vignettes of the R package: https://ghyper.net
Giona's most recent publications
The downside of heterogeneity: How established relations counteract systemic adaptivity in tasks assignments
ArXiv (submitted for publication) - 2021
Configuration models as an urn problem
Scientific Reports 11, 13416 - 2021
Why Online does not Equal Offline: Comparing Online and Real-World Political Support Among Politicians.
socarxiv - 2021
The likelihood-ratio test for multi-edge network models
J. Phys. Complex. 2 035012 - 2021
Fragile, Yet Resilient: Adaptive Decline in a Collaboration Network of Firms
Frontiers in Applied Mathematics and Statistics - 2021
Predicting Sequences of Traversed Nodes in Graphs using Network Models with Multiple Higher Orders
arXiv preprint arXiv:2007.06662 - 2020
Intervention scenarios to enhance knowledge transfer in a network of firms
Frontiers in Physics - 2020
Improving the robustness of online social networks: A simulation approach of network interventions
Frontiers in Robotics and AI - 2020
HYPA: Efficient Detection of Path Anomalies in Time Series Data on Networks
Proceedings of the 2020 SIAM International Conference on Data Mining - 2020
The block-constrained configuration model
Applied Network Science - 2019
Probing the robustness of nested multi-layer networks
A Gaussian Process-based Self-Organizing Incremental Neural Network
2019 International Joint Conference on Neural Networks (IJCNN) - 2019
What is the Entropy of a Social Organization?
Entropy - 2019
Quantifying Triadic Closure in Multi-Edge Social Networks
ACM - 2019
From Relational Data to Graphs: Inferring Significant Links Using Generalized Hypergeometric Ensembles
Social Informatics: 9th International Conference, SocInfo 2017, Oxford, UK, September 13-15, 2017, Proceedings, Part II - 2017