Section: Bilateral Contracts and Grants with Industry

Bilateral Contracts with Industry

• contract with “500px”; PI: Romaric Gaudel.

  Title: Recommender System for Photos

  Duration: May 2016 – Oct. 2016 (6 months)

  Abstract: Recommender Systems aim at recommending items to users. Advances in that field are targeting more and more personalized recommendation. From a recommendation based on market segment to a recommendation based on individual user taste. From a recommendation based on user’s information to a recommendation based on any feedback from any user. From a recommendation based on logged data to a recommendation including latest trends... 500px is a Canadian company which is part of this trend. 500px offers solutions to store pictures online, to share pictures, and to browse among pictures exhibited by other users. Given the huge amount of pictures stored by 500px, users need help to find pictures which corresponds to their tastes. 500px offers several tools to filter the content presented to users. But the tools allowing exploration of the pictures landscape are not personalized, the selection is mostly based on the popularity of pictures/galleries. The most personalized recommendations are obtained by following other users: you see recent pictures of that users. But such recommendations requires you (i) to discover by yourself relevant users, (ii) to explicitly tag these users. The aim of the project is to scan state of the art in Collaborative Filtering and to design a tool which recommends pictures to users based on their implicit actions: given the list of followed users, famed pictures, commented pictures, browsed pictures, ..., infer user’s tastes and recommend to that user pictures and/or other user to look at. The system would also make use of informations on the pictures and of user profiles.

• contract with “Orange Labs”; PI: Philippe Preux

  Title: Sequential Learning and Decision Making under Partial Monitoring

  Duration: Oct. 2014 – Sep. 2017

  Abstract: In applications such as recommendation systems, or computational advertising, the return collected from the user is partial: (s)he clicks on one item, or no item at all. We study this setting in which only a “partial” information is gathered in particular how to learn to behave optimaly in such a setting.

• contract with “55”; PI: Jérémie Mary

  Title: Novel Learning and Exploration-Exploitation Methods for Effective Recommender Systems

  Duration: Oct. 2015 – Sep. 2018

  Abstract: In this Ph.D. thesis we intend to deal with this problem by developing novel and more sophisticated recommendation strategies in which the collection of data and the improvement of the performance are considered as a unique process, where the trade-off between the quality of the data and the performance of the recommendation strategy is optimized over time. This work also consider tensor methods (one layer of the tensor can be the time) with the goal to scale them at RS level.

• contract with “What a nice place” ; PI: Jérémie Mary

  Title: Deduplication of pictures

  Duration: Mar. 2016 – Jan. 2017

  Abstract: “What is nice place” is a start up which aggregates products from different sources in order to provide some home staging advises. Uniqueness of presence for the items in their database can be hard to achieve because of the differences over names and variations of a product. Here we build a classification and deduplication system based on deep neural networks. In this contract we received support from Inria Tech and transferred them some knowledge about deep neural networks.

• contract with “What a nice place” and “Leroy Merlin”; PI: Jérémie Mary

  Title: New Shopping Experience - Virtual Coach

  Duration: Jun. 2016 – Fev. 2017

  Abstract: The goal of this project is to use pictures of house interiors in order to propose automatically some products which would fit in nicely. The relations are learnt automatically using deep neural networks and recommendation systems techniques. We made a first version which focuses on lamps which is available for demonstration at https://whataniceplace.leroymerlin.fr/