Bilateral Self-unbiased Learning from Biased Implicit Feedback

TL;DR

This paper introduces BISER, a novel unbiased recommender system that effectively reduces exposure bias in implicit feedback data by combining self-inverse propensity weighting and bilateral model bridging, leading to improved recommendation accuracy.

Contribution

BISER is the first model to jointly address exposure bias and popularity bias in implicit feedback recommendation systems using a dual-model approach.

Findings

BISER outperforms existing unbiased models on multiple datasets.

The proposed method effectively reduces exposure bias without high computational costs.

Experimental results demonstrate improved recommendation quality across diverse datasets.

Abstract

Implicit feedback has been widely used to build commercial recommender systems. Because observed feedback represents users' click logs, there is a semantic gap between true relevance and observed feedback. More importantly, observed feedback is usually biased towards popular items, thereby overestimating the actual relevance of popular items. Although existing studies have developed unbiased learning methods using inverse propensity weighting (IPW) or causal reasoning, they solely focus on eliminating the popularity bias of items. In this paper, we propose a novel unbiased recommender learning model, namely BIlateral SElf-unbiased Recommender (BISER), to eliminate the exposure bias of items caused by recommender models. Specifically, BISER consists of two key components: (i) self-inverse propensity weighting (SIPW) to gradually mitigate the bias of items without incurring high computational costs; and (ii) bilateral unbiased learning (BU) to bridge the gap between two complementary models in model predictions, i.e., user- and item-based autoencoders, alleviating the high variance of SIPW. Extensive experiments show that BISER consistently outperforms state-of-the-art unbiased recommender models over several datasets, including Coat, Yahoo! R3, MovieLens, and CiteULike.