Description
Date depot: 11 avril 2023
Titre: Unsupervised Contrastive Learning for Attributed Networks
Directeur de thèse:
Mohamed NADIF (Centre Borelli (EDITE))
Encadrant :
Labiod LAZHAR (Centre Borelli (EDITE))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Intelligence artificielle
Resumé: Although, many approaches have emerged with Network Embedding (NE), the research on Attributed Nework Embedding (ANE) is still remains to be explored. Unlike NE that learns from plain networks, ANE aims to capitalize both the proximity information of the network and the affinity of node attributes. In most cases the two tasks representation learning and clustering are taken into account separately, however, there are joint approaches which allow to combine them simultaneously. This reinforces the relationships between these two tasks. The research objective of this thesis project is to contribute to the development of new unified and flexible framework for simultaneous co-clustering and embedding. Under different approaches the benefits of such approach are prooved. Thereby, we intend through this thesis to consider a new joint approach combining contrastive learning and reinforcement learning.