Two papers accepted at KDD 2020! Paper #1: Joint Policy-Value Learning for Recommendation Authors: Olivier Jeunen(intern), David Rohde, Flavian Vasile. Martin Bompaire Abstract. Conventional approaches to recommendation often do not explicitly take into account information on previously shown...
Our REVEAL workshop about Bandit and RL for user interactions will be held at RecSys’20. Co-organizers: Maria Dimakopoulou (Netflix), Thorsten Joachims (Cornell), Olivier Koch (Criteo AI Lab), Yves Raimond (Netflix)Adith Swaminathan (Microsoft), and Flavian Vasile (Crite AI Lab)....
Our workshop will be on either July 17 or 18, and either in Vienna or virtual, depending on the decision of the conference organizers based on the state of the COVID-19 outbreak. The other workshop co-organizers are Michal...
This manuscript introduces the idea of using Distributionally Robust Optimization (DRO) for the Counterfactual Risk Minimization (CRM) problem. Tapping into a rich existing literature, we show that DRO is a principled tool for counterfactual decision making. We also...
tf-yarn is a Python library we have built at Criteo for training TensorFlow models on a YARN cluster. It supports running on one worker or on multiple workers with different distribution strategies, it can run on CPUs or GPUs and...
Professor Noureddine El Karoui (center) with colleagues at Criteo After working on theoretical questions in high-dimensional statistics as an academic at UC, Berkeley, I felt I wanted to see what statistics and machine learning were at industrial scale....
It’s been a year since we launched our AI Lab to drive and shape innovation within the AI field. Among all, we pay attention to give back to the community by sharing our works at top machine learning...
Five Criteos attended on March 27th & 28th the conference ScaledML 2019 at the Computer History Museum in Mountain View. It was the 4th edition of this conference focusing on Scaling ML models, data, algorithms, & infrastructure. The lineup of speakers was impressive:...
At Criteo AI Lab, we are using more and more complex machine learning (ML) models every day. This means that we have many more hyper-parameters to tune. A lot of ML model tuning used to be done by...
