计算机工程2025，Vol.51Issue(5)：103-113,11.DOI:10.19678/j.issn.1000-3428.0069219
基于自适应增强的多视图对比推荐算法
Multi-view Contrastive Recommendation Algorithm Based on Adaptive Enhancement
姚迅 1王海鹏 1胡新荣 1杨捷2
作者信息
- 1. 武汉纺织大学计算机与人工智能学院,湖北武汉 430200
- 2. 伍伦贡大学工程与信息科学学院,澳大利亚伍伦贡2259
- 折叠
摘要
Abstract
Recommendation systems based on neural network architectures have achieved remarkable success in recent years;however,they fail to achieve the desired results when dealing with data rich in popularity biases and interaction noise.Contrastive learning,an emerging technology for learning from unlabeled data,has attracted considerable attention and provides a potential solution to this problem.This study proposes an end-to-end graph-contrastive recommendation method called AMV-CL.This method first constructs a complementary graph of a user-item interaction graph based on the latent representation of nodes and then introduces adaptive augmentation to generate multi-view data from node and edge perspectives.Subsequently,it adjusts the graph structure through a reparameterization network and finally normalizes the sources of positive samples of anchor nodes in contrastive loss,while leveraging multi-view contrastive loss to learn latent representations of users/items.A large number of experiments on public datasets show that,compared with the optimal benchmark method SimGCL,AMV-CL yieldsup to 12.03％and 12.64％improvements in the Recall@20 and NDCG@20 evaluation indicators,respectively.Experimental results show that the proposed method can effectively improve the recommendation performance.关键词
图神经网络/推荐系统/多视图/对比学习/自适应增强Key words
Graph Neural Network(GNN)/recommendation system/multi-view/contrastive learning/adaptive augmentation分类
计算机与自动化引用本文复制引用
姚迅,王海鹏,胡新荣,杨捷..基于自适应增强的多视图对比推荐算法[J].计算机工程,2025,51(5):103-113,11.
