uCTRL: Unbiased Contrastive Representation Learning via Alignment and Uniformity for Collaborative Filtering

TL;DR

uCTRL introduces an unbiased contrastive learning framework for collaborative filtering that effectively mitigates popularity bias and improves recommendation accuracy by optimizing alignment and uniformity functions with a novel IPW method.

Contribution

The paper proposes uCTRL, a contrastive learning approach with unbiased alignment and uniformity functions, enhancing collaborative filtering performance beyond existing unbiased models.

Findings

uCTRL outperforms state-of-the-art models by up to 12.22% in Recall@20.

uCTRL achieves up to 16.33% improvement in NDCG@20.

The method effectively reduces popularity bias in recommendations.

Abstract

Because implicit user feedback for the collaborative filtering (CF) models is biased toward popular items, CF models tend to yield recommendation lists with popularity bias. Previous studies have utilized inverse propensity weighting (IPW) or causal inference to mitigate this problem. However, they solely employ pointwise or pairwise loss functions and neglect to adopt a contrastive loss function for learning meaningful user and item representations. In this paper, we propose Unbiased ConTrastive Representation Learning (uCTRL), optimizing alignment and uniformity functions derived from the InfoNCE loss function for CF models. Specifically, we formulate an unbiased alignment function used in uCTRL. We also devise a novel IPW estimation method that removes the bias of both users and items. Despite its simplicity, uCTRL equipped with existing CF models consistently outperforms state-of-the-art unbiased recommender models, up to 12.22% for Recall@20 and 16.33% for NDCG@20 gains, on four benchmark datasets.