How graph convolutions amplify popularity bias for recommendation?

Graph convolutional networks (GCNs) have become prevalent in recommender system (RS) due to their superiority in modeling collaborative patterns. Although improving the overall accuracy, GCNs unfortunately amplify popularity bias --- tail items are less likely to be recommended. This effect prevents the GCN-based RS from making precise and fair recommendations, decreasing the effectiveness of recommender systems in the long run.

Performance change of LightGCN with different graph convolution layers on Gowalla. Recall@20 and TR@20 stand for the overall recall score and the ratio of tail items in the top-20 recommendation list, respectively.

To solve the problems, a research team led by Xiangnan He (USTC) published their new research on 15 October 2024 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

The team investigated how graph convolutions amplify the popularity bias in RS. Through theoretical analyses, They identified two fundamental factors: (1) with graph convolution i.e., neighborhood aggregation), popular items exert larger influence than tail items on neighbor users, making the users move towards popular items in the representation space; (2) after multiple times of graph convolution, popular items would affect more high-order neighbors and become more influential. The two points make popular items get closer to almost users and thus being recommended more frequently.

To rectify this, the team proposed to estimate the amplified effect of popular nodes on each node's representation, and intervene the effect after each graph convolution. Specifically, they adopted clustering to discover highly-influential nodes and estimate the amplification effect of each node, then remove the effect from the node embeddings at each graph convolution layer. The method is simple and generic --- it can be used in the inference stage to correct existing models rather than training a new model from scratch, and can be applied to various GCN models. They demonstrated the method on two representative GCN backbones LightGCN and UltraGCN, verifying its ability in improving the recommendations of tail items without sacrificing the performance of popular items.

DOI:10.1007/s11704-023-2655-2