Workshops and Tutorials

Abstract

The objective of the course is to give a structured overview of the conceptual frameworks behind the current state-of-the-art recommender systems, explain their underlying assumptions, the resulting methods and their shortcomings and to introduce an exciting new class of approaches, that frames the task of recommendation as a counterfactual policy learning problem. This is a full-day tutorial and it is divided in two modules:In module 1, the participants will learn about the current approaches for building real-world recommender systems, that comprise mainly of two frameworks, namely: recommendation as optimal auto-completion of user behaviour and recommendation as reward modelling. In module 2, we present the framework of recommendation as a counterfactual policy learning problem and go over the theoretical guarantees that address the shortcomings of the previous frameworks. We then proceed to go over the associated algorithms and test them against classical methods in RecoGym, an open-source recommendation simulation environment. Overall, we believe the subject of the course is extremely actual and fills a gap between the consecrated recommendation frame- works and the cutting edge research and sets the stage for future advances in the field. Flavian Vasile is part of the Criteo AI Lab where he works as the ML Recommendations Solutions Architect, with his main focus being on the development of Deep Learning-based Recommendation Systems and on introducing aspects of Causal Inference to Recommendation.Before joining Criteo, he worked as a researcher in the Twitter Advertising Science team; before that, in the Yahoo! Research Lab where he mostly focused on Content Understanding problems. His current research interests include Deep Sequential Models for Recommendation and understanding Recommendation as a decision-making system with reward uncertainty. Among his recent research publications, the work on “Causal Embeddings for Recommendation” received the best paper award at RecSys 2018 and he was the co-organizer of the REVEAL Workshops on Offline Evaluation for Recommender Systems in conjunction with the ACM RecSys conference. David Rohde is a research scientist at Criteo. His research interests are around Bayesian machine learning, offline evaluation and causal inference. He is one of the original creators of the RecoGym environment and one of the authors of the RecoGym course which he delivered at the DS3 summer school in Paris, 2019. He regularly presents at machine learning conferences such as the REVEAL workshop and the causality workshops at NeurIPS 2018 and 2019 Furthermore, he regularly delivers teaching material internally at Criteo and externally at forums such as the DS3 summer school. He has numerous publications in applied and theoretical aspects of machine learning, from topics including variational approximations, causal inference, doubly intractable models to astronomy, analyzing massive public transport datasets and evaluating recommender systems. Olivier Jeunen is a PhD Student at the University of Antwerp, Belgium. His main line of research focuses on implicit-feedback recommender systems, with interests ranging from algorithms and efficiency to evaluation. He regularly collaborates with industrial research labs such as Technicolor, Froomle and Criteo. During this latter collaboration, he contributed to the RecoGym environment and co-authored and co-presented the original RecoGym course multiple times. Additionally, he has several years of experience teaching MSc-level Data Science courses at the University of Antwerp. Amine Benhalloum is a Senior Machine Learning Engineer at Criteo, working on building large scale representation learning and retrieval systems for recommendation, applying Deep learning to personalize billions of daily display ads, reaching billions of users and connecting them with millions of products. His areas of expertise are: large scale machine learning, natural language processing, information retrieval and data intensive systems. Before joining Criteo, Amine worked on a variety of topics ranging from Natural Language processing to fraud detection. He holds a master’s degree in Applied Mathematics.