March 23rd Meetup: Videos are online

The videos from the talks of the second edition of the RecsysFR meetup are now available online.

Slides: http://www.slideshare.net/recsysfr/recommendation-priceministerrakuten-road-to-personalization

Slides: http://www.slideshare.net/recsysfr/rakuten-institute-of-technology-paris

Slides: http://www.slideshare.net/recsysfr/story-of-the-algorithms-behind-deezer-flow

Slides: http://www.slideshare.net/recsysfr/new-tools-from-the-bandit-literature-to-improve-ab-testing

Slides: http://www.slideshare.net/recsysfr/tailormade-personalization-and-recommendation-sailendra

Slides: http://www.slideshare.net/recsysfr/recommendations-at-senscritique-mixing-social-and-machine-learning

March 23rd Meetup: slides are available!

The slides from the talks of the second edition of the RecsysFR meetup are now available online.

Story of the algorithms behind Deezer Flow by Benoit Mathieu & Thomas Boubca, Deezer

Thanks again to the speakers, to our host: PriceMinister and to the audience. We had a great time and hope to see you at the next event.

March 23rd Meetup: Registrations are open

recsysfr_2nd_session

The RecSysFR meetup will be back for a new session on March 23rd at 19:00 in PriceMinister-Rakuten office – 92 Rue Réaumur, 75002 Paris.

The meetup is free, you only have to register on: http://www.meetup.com/fr-FR/RecSysFR/events/229258451/

After an introduction by Patrick Herrmann, PriceMinister CTO – our host for this session, we will have four talks:

• Benoit Mathieu & Thomas Boubca, Deezer“A story of algorithms behind Deezer Flow”
• Emilie Kaufmann, CNRS & CRIStAL: “New tools from the bandit literature to improve A/B Testing”
• Régis Lhoste, Sailendra: “Tailor-made personalization and recommendation”
• Xavier Rampino, SensCritique: “Recommendations at SensCritique – Mixing social and machine learning.”

The talks will be followed by a casual networking time over food and drinks.

NB: only registered attendees will be allowed to participate to the meetup.

 

Dec. 1st meetup: slides are available!

The first recommenders FR meetup was a success with around 50 participants. We had four talks mixing various aspects of recommender systems:

Xavier Dupré (Microsoft, ENSAE) spoke about practical considerations regarding recommender systems in production:

Jérémie Mary (INRIA) presented his work on contextual bandits for recommendation, and gave a wide overview of the current state of the scientific literature on the subject:

Vincent Michel and David Mas (Big Data Europe – PriceMinister – Rakuten group) talked about their current production setup for product recommendation within Rakuten group:

Simon Dollé (Criteo) spoke about the difficulties surrounding product recommendation at scale for display advertising:

 

Thanks a lot to all the speakers for their presentation, we had a great time and we hope to be popping up a second recommenders FR meetup soon!

Dec 1st Meetup: Register now!

The meetup is shaping up and we’ve just open the registration process. Remember that our meetup is free and will happen on Dec 1st from 6:30pm until 9pm at the Criteo office in Paris: 32 rue Blanche, 75009 Paris.

We’ll hear the following talks:

  • Xavier Dupré, Microsoft
  • Jérémie Mary, INRIA: Recommender systems in a sequential context
  • Vincent Michel and David Mas, Big Data Europe / PriceMinister – Rakuten: Recommendations @ Rakuten Group
  • Pierre-Emmanuel Mazaré, Criteo: Large scale recommendation for display advertising

The talks will be followed by a casual networking time over food and drinks. To register, just use the box below or go to EventBrite.

Criteo at ACM RecSys 2015 in Vienna

A couple of weeks ago, Criteo attended the ACM Recsys 2015 conference at the Technische Univerität Wien, Vienna.

Criteo makes display perform better : display the right ad, at the right moment, to the right user. So Romain Lerallut and Diane Gasselin gave a talk about the large scale real-time product recommender we use: what we do, how we do it, and what are the new challenges.

We must process recommendation requests in less than 100ms while we juggle with 1B users, 3B products large data sources and 20B/day of display data. In order to scale and be reactive, all the products cannot be filtered and ranked online. To address this issue, we pre-select sources of product offline using collaborative-filtering and logistic regression scoring and we then refine the score with the ad context online.

We are also facing various new challenges: offline ab-testing, longer-term user profile, NLP features, instant updates, full banner scoring, …


by Diane Gasselin for the Criteo R&D team.