Overview
Network models are widely used to study complex systems, but many assume that connections are independent once the communities of the nodes are known. This assumption can be unrealistic: for example, one directed interaction may make a reciprocal interaction more likely.
Led by Caterina De Bacco, this project—Relaxing Restrictive Independence Assumptions in Networks—developed probabilistic generative models that better represent dependencies between network connections. The project examined multilayer networks, reciprocal interactions, dynamic networks, and higher-order networks in which several entities can interact simultaneously. It concluded in December 2023.
Key results
- Developed probabilistic network models that jointly infer community structure and dependencies between edges.
- Introduced a model based on a bivariate Bernoulli distribution to analyse pairs of reciprocal edges directly, rather than treating them as independent.
- Integrated node attributes and multiple relationship types into community-detection models, improving tasks such as predicting missing links or attributes.
- Extended the methods to dynamic networks, where connections appear and disappear over time, and to hypergraphs, which represent group interactions involving more than two nodes.
- Developed methods for inferring missing hyperedges and detecting overlapping communities. These approaches performed well on synthetic and real-world data and scaled to very large hypergraphs.
- Applied the models to noisy social-network data in which relationships were reported by multiple people. The models estimate the underlying network while accounting for differences in how accurately individuals report connections.
- Contributed to HypergraphX, an open-source Python library for analysing higher-order networks.
Why it matters
Many real-world systems involve dependent, reciprocal, multilayer, or group-based interactions that simpler network models cannot represent adequately. By relaxing restrictive independence assumptions, this research enables more realistic analyses of social, technological, biological, and communication networks. The methods can support community detection, link prediction, anomaly detection, and the study of higher-order relationships.
Publications and output
- M. Contisciani, E. A. Power, and C. De Bacco. “Community detection with node attributes in multilayer networks”. Scientific Reports 10.1 (2020), page 15736.
- K.Higham, M. Contisciani, and C. De Bacco. “Multilayer patent citation networks: A comprehensive analytical framework for studying explicit technological relationships”. Technological Forecasting and Social Change 179 (2022), page 121628.
- H. Safdari, M. Contisciani,and C. De Bacco. “Generative model for reciprocity and community detection in networks”. Physical Review Research 2 (2021).
- M. Contisciani, H. Safdari, and C. De Bacco. “Community detection and reciprocity in networks by jointly modelling pairs of edges”. Journal of Complex Networks 10.4 (2022).
- H. Safdari, M. Contisciani, and C. De Bacco. “Reciprocity, community detection, and link prediction in dynamic networks”. Journal of Physics: Complexity 3.1 (2022).
- C. De Bacco, M. Contisciani, J. Cardoso-Silva, H. Safdari, G. Lima Borges, D. Baptista, T. Sweet, J.-G. Young, J. Koster, C. T. Ross, R. McElreath, D. Redhead, and E. A. Power. “Latent network models to account for noisy, multiply reported social network data”. Journal of the Royal Statistical Society Series A: Statistics in Society 183 (2023).
- H. Safdari, M. Contisciani, and C. De Bacco. “Anomaly, reciprocity, and community detection in networks”. Physical Review Research 5.3 (2023).
- M. Contisciani, F. Battiston, and C. De Bacco. “Inference of hyperedges and overlapping communities in hypergraphs”. Nature Communications 13.1 (2022).
- N. Ruggeri, M. Contisciani, F. Battiston, and C. De Bacco. “Community detection in large hypergraphs”. Science Advances 28 (2023).
- Q. F. Lotito, M. Contisciani, C. De Bacco, L. Di Gaetano, L. Gallo, A. Montresor, F. Musciotto, N. Ruggeri, and F. Battiston. “Hypergraphx: a library for higher-order network analysis”. Journal of Complex Networks 11.3 (2023).
