图卷积或者图神经网络适合做推荐系统吗?
关注者
401被浏览
278,970登录后你可以
不限量看优质回答私信答主深度交流精彩内容一键收藏
这方面论文挺多的啊。
你想,用户-项目可以构成二部图(Bipartite Graph),用户与用户之间可以构成社交网络(Social Network),项目与项目之间可以存在知识图谱(Knowledge Graph),另外把这几者都考虑进去可以构成异质图(Heterogeneous Information Network),再把时间因素考虑进来而产生的动态演化而构成动态图(Dynamic Graph)。
所以推荐系统中许多形式的数据都可以表示成图,当然许多大佬自然而然的将强大的GNN应用到推荐领域了,以下列举几篇文献,可以看看。
- Graph Convolutional Matrix Completion. 2017.
- Graph Convolutional Neural Networks for Web-Scale Recommender Systems. KDD 2018.
- Graph Contextualized Self-Attention Network for Session-based Recommendation. IJCAI 2019.
- Session-based Recommendation with Graph Neural Networks. AAAI 2019.
- Knowledge-aware Graph Neural Networks with Label Smoothness Regularization for Recommender Systems. KDD 2019.
- KGAT: Knowledge Graph Attention Network for Recommendation. KDD 2019.
- Knowledge Graph Convolutional Networks for Recommender Systems. WWW 2019.
- Graph Neural Networks for Social Recommendation. WWW 2019.
- Memory Augmented Graph Neural Networks for Sequential Recommendation. AAAI 2020.
- Revisiting Graph based Collaborative Filtering: A Linear Residual Graph Convolutional Network Approach. AAAI 2020.