profile-pic
LargeRecSys
The workshop for large scale recommender systems, held in conjunction with the 2024 ACM conference on Recommender Systems.

Introduction


Recommender systems are ubiquitous across many domains such as entertainment, e-commerce, news, social media, fitness, e-commerce and education, and as such, the need to think about the problems at scale is ever growing. As the Scaling Law in Recommender Systems becomes more apparent, the demand for billion-parameter or even trillion-parameter models in Recommender Systems is increasing to unprecedented levels. When implemented within industry settings, these systems encounter many real-world constraints and complexities that require tailored solutions. Designing and maintaining recommender systems that operate at a massive scale come with their own set of distinct algorithmic, systematic, and ethical challenges.

LargeRecSys workshop aims to bridge the gap between state-of-the-art RecSys research and the practical demands of operating large recommender systems at scale. We plan to invite researchers and practitioners from both industry and academia to engage in discussions and challenges inherent in large-scale recommender systems.

Call For Papers


We invite submissions in two formats: extended abstracts (1-8 pages), or slides (15-20 slides). By accepting slides, we hope to lower the writing burden for industry participants. However, since slides submissions sometimes are short on details, we might request clarification or additional editing as condition for acceptance. We encourage contributions in new theoretical research, practical solutions to particular aspects of scaling a recommender, best practices in scaling evaluation systems, and creative new applications of big data to large scale recommendation systems.

Our topics of interests include, but are not limited to:

  • Large Foundational Models and Large User Representations: Exploring techniques in large user-modeling and learning large foundational models that adapt well to highly dynamic environments and scale well with content and users. Enabling adaptive and real-time learning and optimization of recommendations in the face of fast evolving user-preferences and content uploads.
  • Large Language Models (LLM) for Recommendations: Exploring the utilization of LLM to enhance recommendation systems, leveraging natural language understanding to generate personalized suggestions across various content domains.
  • Scalability and Efficiency: Investigating and discussing system architectures and optimizations that effectively process billions of interactions, and ensure efficiency in model training, serving and updates.
  • Data Sparsity and Cold-start: Developing data-efficient methods to deal with the lack of user or item data to improve large-scale recommendations in data sparse environments.
  • Fairness and Bias in Large Ecosystems: Designing large recommender systems that address issues of algorithmic bias, and promote diversity and fairness.
  • Evaluation and and Experimentation: Establishing online and offline metrics and algorithms to measure user satisfaction and ecosystem objectives, going far beyond the classic offline accuracy-based measures. Discussing offline and online iteration and experimentation frameworks to test at scale efficiently and reduce data pollution and leakage.
  • RecSys MLOps: Streamlining model deployment, monitoring and retraining to ensure the performance and stability of production-scale recommender systems.
Add Submission

Keynote Speakers


Lexi Baugher

Lexi Baugher

Lexi Baugher is a Distinguished Software Engineer at YouTube. She got her Bachelor of Science from Caltech in Engineering and Applied Science. She has worked at Google for 23 years. She is currently the Area Tech Lead of YouTube Discovery, overseeing development of YouTube's recommendation systems. She lives in San Francisco with her husband, kid, and cat; and enjoys baseball, running and baking.

Justin Basilico

Justin Basilico

Justin Basilico is a research/engineering director at Netflix. He leads the recommendations algorithms team that does research and development of the machine learning algorithms used to power the core Netflix experience and personalize it for our members worldwide. Prior to Netflix, he worked in the Cognitive Systems group at Sandia National Laboratories. He has an MS in Computer Science from Brown University and a BA in Computer Science from Pomona College.

Important Dates


Paper Submission Deadline: August 30, 2024

Reviewers Deadline: September 11, 2024

Author Notification: September 13, 2024

Camera-Ready Deadline: September 20, 2024

Organizers