Lightweight representation learning for efficient and scalable recommendation

TL;DR

This paper introduces LED, a lightweight recommendation model that balances complexity, scale, and performance, enabling real-time recommendations for billions of users and items using standard hardware.

Contribution

The paper presents a novel lightweight encoder-decoder model that combines large-scale matrix factorization with embedding fine-tuning for scalable recommendation systems.

Findings

Achieves state-of-the-art performance at internet scale

Serves billions of users and millions of items in milliseconds

Operates reliably over two months in a real-world setting

Abstract

Over the past decades, recommendation has become a critical component of many online services such as media streaming and e-commerce. Recent advances in algorithms, evaluation methods and datasets have led to continuous improvements of the state-of-the-art. However, much work remains to be done to make these methods scale to the size of the internet. Online advertising offers a unique testbed for recommendation at scale. Every day, billions of users interact with millions of products in real-time. Systems addressing this scenario must work reliably at scale. We propose an efficient model (LED, for Lightweight Encoder-Decoder) reaching a new trade-off between complexity, scale and performance. Specifically, we show that combining large-scale matrix factorization with lightweight embedding fine-tuning unlocks state-of-the-art performance at scale. We further provide the detailed description of a system architecture and demonstrate its operation over two months at the scale of the internet. Our design allows serving billions of users across hundreds of millions of items in a few milliseconds using standard hardware.