Conférence
Notice
Lieu de réalisation
Centre Inria d'Université Côte d'Azur
Langue :
Français
Crédits
Nelly Litvak (Intervention)
Crédit image : Centre Inria d'Université Côte d'Azur
Détenteur des droits
Centre Inria d'Université Côte d'Azur
Conditions d'utilisation
Droit commun de la propriété intellectuelle
DOI : 10.60527/7mkq-nk51
Citer cette ressource :
Nelly Litvak. Inria. (2023, 5 décembre). Projection methods for community detection in complex networks. [Vidéo]. Canal-U. https://doi.org/10.60527/7mkq-nk51. (Consultée le 25 avril 2025)
logo Inria
Inria - Institut national de recherche en sciences et technologies du numérique
S'abonner
Contacter

Projection methods for community detection in complex networks

Réalisation : 5 décembre 2023 - Mise en ligne : 14 décembre 2023
  • document 1 document 2 document 3
  • niveau 1 niveau 2 niveau 3
Descriptif

In this work we present a new framework that rephrases the task of community detection as a nearest neighbor search on a high-dimensional hypersphere. This yields a new class of community detection methods, which we call the `projection methods’. A projection method consists of two steps: 1) The networks is mapped to a query vector on a high-dimensional hypersphere. 2) The query vector is projected to the set of clusterings, where each clustering, too, is a vector on the hypersphere; this projection step is done by minimizing the distance between the query vector and the clustering, over the set of all clusterings. We prove that projection methods generalize many popular community detection methods, including modularity maximization. This has many practical implications. For example, modularity maximization and many other methods suffer from the same `granularity problem’: they have some parameter that controls granularity of the obtained clustering, but no method to predict the obtained granularity from the value of this parameter. In projection methods, the granularity of the obtained clustering is close to that of the query vector. With this insight, we provide a general heuristic for choosing a query vector to address the granularity problem. Another advantage of the projection methods is that a query vector may explicitly depend on various network characteristics (edges, wedges, triangles), thus seamlessly including these in community detection.

Intervention
Thème
Discipline :

Sur le même thème