Optimizing top-n collaborative filtering via dynamic negative item sampling

TLDR

Collaborative filtering relies on aggregated user preference data, but implicit feedback only reveals positive interactions, leaving negative examples unobserved, and prior methods that randomly sample negatives or use heuristics introduce bias and slow training. This work proposes to dynamically select negative training samples from the ranked list generated by the current model and iteratively update the model. The method repeatedly draws negatives from the model’s own ranking and retrains, thereby aligning negative sampling with evolving predictions. Experiments on three large‑scale datasets demonstrate that this dynamic sampling shortens training time and yields significant performance improvements.

Abstract

Collaborative filtering techniques rely on aggregated user preference data to make personalized predictions. In many cases, users are reluctant to explicitly express their preferences and many recommender systems have to infer them from implicit user behaviors, such as clicking a link in a webpage or playing a music track. The clicks and the plays are good for indicating the items a user liked (i.e., positive training examples), but the items a user did not like (negative training examples) are not directly observed. Previous approaches either randomly pick negative training samples from unseen items or incorporate some heuristics into the learning model, leading to a biased solution and a prolonged training period. In this paper, we propose to dynamically choose negative training samples from the ranked list produced by the current prediction model and iteratively update our model. The experiments conducted on three large-scale datasets show that our approach not only reduces the training time, but also leads to significant performance gains.