SimCE: Simplifying Cross-Entropy Loss for Collaborative Filtering

TL;DR

This paper introduces SimCE, a simplified and more effective loss function for collaborative filtering that leverages multiple negative samples, leading to improved convergence and performance over existing methods like BPR and SSM.

Contribution

The paper proposes SimCE, a simplified upper bound of SSM, which enhances training efficiency and recommendation accuracy in collaborative filtering systems.

Findings

SimCE outperforms BPR and SSM on 12 benchmark datasets.

Using multiple negative samples improves recommender system performance.

SimCE demonstrates better convergence and results with MF and LightGCN backbones.

Abstract

The learning objective is integral to collaborative filtering systems, where the Bayesian Personalized Ranking (BPR) loss is widely used for learning informative backbones. However, BPR often experiences slow convergence and suboptimal local optima, partially because it only considers one negative item for each positive item, neglecting the potential impacts of other unobserved items. To address this issue, the recently proposed Sampled Softmax Cross-Entropy (SSM) compares one positive sample with multiple negative samples, leading to better performance. Our comprehensive experiments confirm that recommender systems consistently benefit from multiple negative samples during training. Furthermore, we introduce a \underline{Sim}plified Sampled Softmax \underline{C}ross-\underline{E}ntropy Loss (SimCE), which simplifies the SSM using its upper bound. Our validation on 12 benchmark datasets, using both MF and LightGCN backbones, shows that SimCE significantly outperforms both BPR and SSM.