Context

Recommender systems emerged as an independent research area in the mid-1990s when researchers started focusing on recommendation problems that explicitly rely on the ratings structure. In its most common formulation, the recommendation problem is reduced to the problem of estimating ratings for the items that have not been seen by a user. Meanwhile, thanks to the development of many social services like tagging, bookmarking and friendship, more and more data on social interaction becomes available. All these new sources of heterogeneous data can be represented as multiple graphs, linking users and items. One challenge is therefore to combine efficiently these data sources to obtain more reliable recommender systems.

Proposed Research

The main approach will be to consider recommendation as ranking on a

(multiple) graph(s) problem. One aspect of this project proposal is to leverage existing approaches to the case of heterogeneous data by means of multiple graphs combination techniques.

1. State of the art concerning graph based approaches in recommender systems.
2. Acquisition of datasets from popular websites in the topic of music recommendation and scientific paper recommendation.
3. Proposition of multiple graphs and hypergraphs combination for ranking.
4. Implementation and experimentation

Good skills in machine learning, algorithms and statistics are required. Upon successful completion of this internship a PhD student position might be considered.

Contact

The intership will take place inside the Magnet team of the INRIA Lille research center. Magnet was part of Mostrare Project-Team and aims to develop learning algorithms in the presence of graph structured data.

Mikaela Keller, Pascal Denis and Marc Tommasi: first.last@inria.fr

Bibliography