Stochastic block models for random graphs analysis
- Mathématiques et informatique appliquées aux sciences humaines et sociales
- 2026
- 57 min 16 s
- Anglais
Publié le 21/07/2017
Stochastic block models are latent variable models where the nodes of a random graph are assumed to belong to unobserved groups. Conditional on those nodes groups, edges between two nodes are independent random variables. Moreover, the conditional distribution of an edge depends on the groups of the nodes that it links. These models thus perform a probabilistic clustering of the graph nodes. We will discuss recent results on stochastic block models (SBM) as well as some generalizations, such as overlapping SBM or latent block models for array data.
Contribution
Université Paris 1 Panthéon - Sorbonne
Catherine Matias
