Prototypical Contrastive Learning and Adaptive Interest Selection for Candidate Generation in Recommendations

TL;DR

This paper introduces an adaptive interest selection mechanism and a prototypical contrastive learning approach to improve candidate generation in large-scale recommender systems, addressing interest diversity and negative sampling issues.

Contribution

It proposes a novel adaptive interest selection layer and a prototypical contrastive learning module to enhance user interest modeling and negative sample handling.

Findings

Significant performance improvements over baselines on multiple benchmarks.

Effective adaptation of user interest embeddings based on user activeness.

Reduction of class collision issues in negative sampling.

Abstract

Deep Candidate Generation plays an important role in large-scale recommender systems. It takes user history behaviors as inputs and learns user and item latent embeddings for candidate generation. In the literature, conventional methods suffer from two problems. First, a user has multiple embeddings to reflect various interests, and such number is fixed. However, taking into account different levels of user activeness, a fixed number of interest embeddings is sub-optimal. For example, for less active users, they may need fewer embeddings to represent their interests compared to active users. Second, the negative samples are often generated by strategies with unobserved supervision, and similar items could have different labels. Such a problem is termed as class collision. In this paper, we aim to advance the typical two-tower DNN candidate generation model. Specifically, an Adaptive Interest Selection Layer is designed to learn the number of user embeddings adaptively in an end-to-end way, according to the level of their activeness. Furthermore, we propose a Prototypical Contrastive Learning Module to tackle the class collision problem introduced by negative sampling. Extensive experimental evaluations show that the proposed scheme remarkably outperforms competitive baselines on multiple benchmarks.