A0680
Title: Partially exchangeable stochastic block models for multilayer networks
Authors: Francesco Gaffi - University of Notre Dame (United States) [presenting]
Daniele Durante - Bocconi University (Italy)
Antonio Lijoi - Bocconi University (Italy)
Igor Pruenster - Bocconi University (Italy)
Abstract: There is an increasing availability of multilayer network data but still a lack of state-of-the-art models for node-coloured multilayer networks, which can flexibly account for both within and across-layer block-connectivity structures while incorporating layer information in a principled probabilistic manner. Such a gap is covered by proposing a new class of partially exchangeable stochastic block models that relies on a hierarchical random partition prior to the group allocations of nodes driven by the urn scheme of a hierarchical normalized completely random measure. The partial exchangeability assumption among nodes according to layer partitions allows inferring both within- and across-layer blocks while preserving probabilistic coherence, principled uncertainty quantification and formal inclusion of prior information from the layer division. The mathematical tractability and projectivity of the construction further allow analytically deriving predictive within- and across-layer co-clustering probabilities, thus facilitating prior elicitation and development of rigorous predictive strategies for both the connections and allocations of future incoming nodes. The practical performance of this novel class is illustrated in simulation studies and in a real-world criminal network application, where the proposed model displays clear gains relative to alternative solutions in estimation, uncertainty quantification and prediction.
