1 Title
Self-Supervised Hypergraph Convolutional Networks for Session-based Recommendation(Xin Xia,Hongzhi Yin,Junliang Yu,Qinyong Wang,Lizhen Cui,Xiangliang Zhang)【AAAI 2021】
2 Conclusion
This papers fill the gap that SBR methods regard the item transitions as pairwise relations, which neglect the complex high-order information among items and hypergraph provides a natural way to capture beyond-pairwise relations, while its potential for SBR has remained unexplored by modeling sessionbased data as a hypergraph and then propose a hypergraph convolutional network to improve SBR. Moreover, to enhance hypergraph modeling, this study devises another graph convolutional network which is based on the line graph of the hypergraph and then integrate self-supervised learning into the training of the networks by maximizing mutual information between the session representations learned via the two networks, serving as an auxiliary task to improve the recommendation task.
3 Good Sentences
1、In RNNs-based approaches, modeling session-based data as unidirectional sequences is deemed as the key to success, since the data is usually generated in a short period of time
and is likely to be temporally dependent. However, this assumption may also trap these RNNs based models because it ignores the coherence of items. AIn RNNs-based approaches, modeling session-based data as unidirectional sequences is deemed as the key to success,
since the data is usually generated in a short period of time and is likely to be temporally dependent. However, this assumption may also trap these RNNs-based models because it ignores the coherence of items. (The shortcomings of RNNs-based models which were used in Recommendation systems)
2、Besides, the high complexity of this model makes it impossible to be deployed in real scenarios. Currently, there is no research bridging hypergraph neural networks and SBR, and we are the first to fill this gap.(The creativoty of this study had done)
3、The hypergraph modeling empowers our model to achieve significant performance. However, we consider that the sparsity of session data might impede hypergraph modeling, which would result in a suboptimal recommendation performance. Inspired by the successful practices of selfsupervised learning on simple graphs, we innovatively integrate self-supervised learning into the network to enhance hypergraph modeling(The usage of Self-Supervised Learning in the Hypergraph Recommender systems)
本文提出一种双通道超图推荐系统的应用。