关键词: 图协同过滤, 过平滑, 自我特征, 对比学习, 图神经网络, 个性化推荐

Abstract: Aiming at the problems of over-smoothing and noise in the embedding representation in the message passing process of graph convolution based on graph neural network recommendation, a Recommendation model combining Self-features and Contrastive Learning (SfCLRec) were proposed. The model was trained using a pre-training-formal training architecture. Firstly, the embedding representation of users and items was pretrained to maintain the feature uniqueness of the nodes themselves by fusing the node self-features and introducing a hierarchical contrastive learning task to mitigate the noisy information from the higher-order neighboring nodes. Then, the collaborative graph adjacency matrix was reconstructed according to the scoring mechanism in the formal training stage. Finally, the predicted score was obtained based on the final embedding. Compared with existing graph neural network recommendation models such as LightGCN and SimGCL, SfCLRec achieves the better recall and NDCG (Normalized Discounted Cumulative Gain) in three public datasets ML-latest-small, Last.FM and Yelp, validating the effectiveness of the model.

Key words: graph collaborative filtering, over-smoothing, self-feature, contrastive learning, graph neural network, personalized recommendations