Online Adaptive Recommendation in Streaming Environments
AS6 Level 5
MR6, AS6-05-10
Abstract:
Most recommender systems proposed in the literature build first a model from a large static dataset, and then rebuild it periodically as new chunks of data arrive and are added to the original dataset. This functioning in batch faces several limitations, and with the explosion of the volume of user-generated data, it has become essential to design online recommender systems that continuously learn from data streams and that adapt to changes in real-time.
This talk will focus on the topic of stream-based recommender systems. In the first part, we will present the characteristics and requirements of such systems with respect to different application scenarios. We will also highlight the general trends and directions of existing work. In the second part, we will focus on the problem of handling concept drifts in stream-based recommender systems, while introducing a general framework for online adaptive recommendation. We will present and discuss two approaches developed for this purpose, namely adaptive incremental matrix factorization and adaptive collaborative topic modeling, where the goal is to adapt to drifts in items' perceptions and descriptions. Finally, we will mention general directions for future research on this topic.
Biodata:
Marie Al-Ghossein is a postdoctoral researcher at Telecom Paris, France, in the DIG team. She received her Ph.D. from Telecom Paris in 2019 and was working under the supervision of Pr. Talel Abdessalem and Anthony Barre, while also being part of the Data Science team of AccorHotels in Paris. Her research interests lie in the fields of data mining and machine learning and their application to recommender systems and content personalization. Specifically, she has mainly worked on the problem of recommendation in the tourism domain and also on stream-based recommender systems. Prior to doing her Ph.D., she received a double degree in Engineering from Telecom Paris and Saint Joseph University in Beirut, Lebanon.