PhD defense of Chakib FETTAL :

Chakib FETTAL

Contributions to Scalable Clustering of Networks and Graphs

Abstract

This thesis tries to address scalability issues of the state-of-the-art graph clustering models and presents approaches for clustering and representation learning different types of graphs, including classical graphs, bipartite graphs, attributed graphs, bipartite attributed graphs, and multi-view attributed graphs. To this end we leverage techniques such as : linear projections, Laplacian smoothing, optimal transport, etc. The proposed approaches all share three key characteristics: simplicity, cost-effectiveness, and having few hyper-parameters. Thanks to their simple yet effective nature, the proposed methods are competitive with the state of the art while also generally being more computationally efficient. We showcase the efficacy and efficiency of our models against state-of-the-art methods through extensive experimentation and significance testing.

Supervision

• Mohamed Nadif
• Lazhar Labiod

Jury

• M. Marsala Christophe, Professeur des universités, LIP6, Sorbonne Université
• Mme. Niang Ndeye, Professeure des universités, Centre d’études et de recherche en informatique et communication, CNAM
• M. Adam Sébastien     Professeur des universités     Laboratoire d'informatique de traitement de l'information et des systèmes, Univ. Rouen
• M. Lenca Philippe     Professeur des universités     Laboratoire des sciences et techniques de l'information, de la communication et de la connaissance
• M. Nadif Mohamed     Professeur des universités     Centre Borelli (EDITE), Univ. de Paris Cité