Unbiased Knowledge Distillation for Recommendation

TL;DR

This paper introduces an unbiased knowledge distillation method for recommender systems that mitigates popularity bias by stratifying items during the distillation process, leading to fairer and more accurate recommendations.

Contribution

It proposes a stratified distillation strategy that reduces bias in knowledge transfer, improving recommendation fairness without altering teacher training.

Findings

Reduces popularity bias in distilled recommender models

Improves recommendation fairness and accuracy

Validated through extensive theoretical and empirical studies

Abstract

As a promising solution for model compression, knowledge distillation (KD) has been applied in recommender systems (RS) to reduce inference latency. Traditional solutions first train a full teacher model from the training data, and then transfer its knowledge (\ie \textit{soft labels}) to supervise the learning of a compact student model. However, we find such a standard distillation paradigm would incur serious bias issue -- popular items are more heavily recommended after the distillation. This effect prevents the student model from making accurate and fair recommendations, decreasing the effectiveness of RS. In this work, we identify the origin of the bias in KD -- it roots in the biased soft labels from the teacher, and is further propagated and intensified during the distillation. To rectify this, we propose a new KD method with a stratified distillation strategy. It first partitions items into multiple groups according to their popularity, and then extracts the ranking knowledge within each group to supervise the learning of the student. Our method is simple and teacher-agnostic -- it works on distillation stage without affecting the training of the teacher model. We conduct extensive theoretical and empirical studies to validate the effectiveness of our proposal. We release our code at: https://github.com/chengang95/UnKD.