Live Session
Teatro Petruzzelli
Paper
16 Oct
 
12:00
CEST
Session 9: Sequential Recommendation 2
Add Session to Calendar 2024-10-16 12:00 pm 2024-10-16 01:10 pm Europe/Rome Session 9: Sequential Recommendation 2 Session 9: Sequential Recommendation 2 is taking place on the RecSys Hub. Https://recsyshub.org
Main Track

Scalable Cross-Entropy Loss for Sequential Recommendations with Large Item Catalogs

View on ACM Digital Library

Gleb Mezentsev (Skoltech), Danil Gusak (Skoltech, HSE), Ivan Oseledets (AIRI, Skoltech) and Evgeny Frolov (AIRI, HSE, Skoltech)

View Paper PDFView Poster
Abstract

Scalability issue plays a crucial role in productionizing modern recommender systems. Even lightweight architectures may suffer from high computational overload due to intermediate calculations, limiting their practicality in real-world applications. Specifically, applying full Cross-Entropy (CE) loss often yields state-of-the-art performance in terms of recommendations quality. Still, it suffers from excessive GPU memory utilization when dealing with large item catalogs. This paper introduces a novel Scalable Cross-Entropy (SCE) loss function in the sequential learning setup. It approximates the CE loss for datasets with large-size catalogs, enhancing both time efficiency and memory usage without compromising recommendations quality. Unlike traditional negative sampling methods, our approach utilizes a selective GPU-efficient computation strategy, focusing on the most informative elements of the catalog, particularly those most likely to be false positives. This is achieved by approximating the softmax distribution over a subset of the model outputs through the maximum inner product search. Experimental results on multiple datasets demonstrate the effectiveness of SCE in reducing peak memory usage by a factor of up to 100 compared to the alternatives, retaining or even exceeding their metrics values. The proposed approach also opens new perspectives for large-scale developments in different domains, such as large language models.

Join the Conversation

Head to Slido and select the paper's assigned session to join the live discussion.

Conference Agenda

View Full Agenda →
No items found.