Recommender Systems: Online Learning and Ranking

Type de soutenance
Date de début
Date de fin
IRISA Rennes
Départements supplémentaires
Recommender Systems are recommending items to users, choosing the items in a list a thousands of potential one and aiming at those which will please the users. They are a crucial ingredient for websites recommending movies, songs, books, products...

The recommendation is done based on a model of the preferences of users w.r.t. items, where the parameters of the model are learned from users’ feedback, such as listening time, clicks, or rates. This feedback reflects users’ appreciation with respect to each displayed item.
I will present an overview of my contributions towards Recommender Systems, with a focus on two practical aspects: the bandit aspect of the feedback, and the link with Learning to Rank.

Composition du jury
Marianne CLAUSEL - Professor, Univ. Lorraine, France - Reviewer
Tor LATTIMORE - Senior Research Scientist at Deepmind - Reviewer
Liva RALAIVOLA - Head of AI Research at Criteo AI Lab, France (on leave from Professor at Univ. Marseille) - Reviewer
Massih-Reza AMINI - Professor, Univ. Grenoble Alpes, France
Élisa FROMONT - Professor, Univ Rennes, France