Ratings and comparisons for recommender systems

Contact: Oscar Villemaud

We are in an era of content overload. There is far too much content available online for humans to sort through it individually. This is why most of what we see on the web is picked by recommendation algorithms. In this project, you will study the effectiveness of recommendation algorithms and think about how to improve them. In particular, you will focus on which types of questions the users should be asked in order to undersand their preferences efficiently and accurately. The project can be theoretical or experimental.

[1] Mnih, Andriy, and Russ R. Salakhutdinov. "Probabilistic matrix factorization." Advances in neural information processing systems 20 (2007).

[2] Maystre, Lucas. Efficient learning from comparisons. No. THESIS. EPFL, 2018. (Introduction)

[IMPORTANT : no availability for bachelor students anymore]

Requirements :

Python , pytorch is a plus

Basis of linear algebra