Connected Content retriever boosts LinkedIn feed engagement with GPU power
Connected Content Retriever: Dense Graph Edge Features Powering Pre-Ranking at LinkedIn
Machine Learning
Summary
Showing the most interesting posts on a social media feed is hard because there are millions of possible posts to choose from. The authors built a system called Connected Content Retriever that uses powerful graphics processors (GPUs) to quickly score and pick the best few hundred posts based on connections and interactions within LinkedIn’s network. This system uses detailed relationship features to understand how closely content relates to the viewer, even through friends of friends. By doing this, LinkedIn was able to improve the time users spent on their feed by 2.5%, which is a significant gain.
What this means in practice
- •For social media platform engineers: Implement low-latency GPU-based pre-ranking to efficiently filter large candidate sets using graph features for improved content recommendations.
- •For e-commerce personalizers: Use dense graph edge features and GPU scoring for fast pre-ranking of product recommendations leveraging user-item interaction networks.
Authors
Akhilesh Gupta, Sudarshan Srinivasa Ramanujam, Chirag Bhanuprasad Mehta, Reshma Asharaf Beena, Dhritiman Das, Birjodh Singh Tiwana, Bhargavkumar Kanubhai Patel, Mack Lee, Renyi Tang
Abstract
In large-scale recommendation systems like the LinkedIn Feed, content generated by a member's network (connections and follows) makes up over 70% of impressions and engagement. It is therefore essential that the pre-ranking layer forwards the best possible few hundred candidates to the ranking layer. LinkedIn's professional knowledge graph carries engagement signals across both the first degree network (connections and follows) and the second-degree network: posts that a 1st-degree connection reacted to, commented on or reshared but did not author (a.k.a. stranger viral). Due to this fan out, the resulting candidate index exceeds one billion; selection of activities from the viewer's network narrows it down to roughly tens of thousands of activities that must be scored within a 120 ms p99 latency budget. We present Connected Content Retriever (CC Retriever), a pre-ranking system that scores these candidates with a full deep ranking model on GPUs at low latency. At its core is a sorted-search GPU primitive that joins dense graph affinity features (viewer to author) with document level features stored on the GPU at runtime in 5-10 ms. The shift to GPU served scoring enabled a 50x scale up of the ranking model's parameters and delivered a +2.5% lift in content time spent on the LinkedIn Feed in online experiments, significantly higher than the typical gains observed in LinkedIn Feed experiments. In this work, we describe the feature set we leverage from LinkedIn's economic graph and the model architecture used for scoring, with a particular emphasis on the online system that scales the stack.