Revisiting Recommendation Loss Functions through Contrastive Learning (Technical Report)

TL;DR

This paper systematically examines recommendation loss functions through contrastive learning, introducing new variants like InfoNCE+ and MINE+ that improve recommendation performance, and reveals inherent debiasing in some linear models.

Contribution

It introduces InfoNCE+ and MINE+ as optimized contrastive loss variants, and proposes Debiased CCL to improve recommendation accuracy and reduce bias.

Findings

MINE+ outperforms existing loss functions in recommendation tasks.

Debiased CCL effectively reduces bias in pointwise recommendation loss.

Linear models like iALS and EASE are inherently debiased.

Abstract

Inspired by the success of contrastive learning, we systematically examine recommendation losses, including listwise (softmax), pairwise (BPR), and pointwise (MSE and CCL) losses. In this endeavor, we introduce InfoNCE+, an optimized generalization of InfoNCE with balance coefficients, and highlight its performance advantages, particularly when aligned with our new decoupled contrastive loss, MINE+. We also leverage debiased InfoNCE to debias pointwise recommendation loss (CCL) as Debiased CCL. Interestingly, our analysis reveals that linear models like iALS and EASE are inherently debiased. Empirical results demonstrates the effectiveness of MINE+ and Debiased-CCL.