Dual Side Deep Context-aware Modulation for Social Recommendation

Cited by: 0|Views6
Weibo:
To address the above issues, we proposed DICER

Abstract:

Social recommendation is effective in improving the recommendation performance by leveraging social relations from online social networking platforms. Social relations among users provide friends' information for modeling users' interest in candidate items and help items expose to potential consumers (i.e., item attraction). However, th...More

Code:

Data:

0
Full Text
Bibtex
Weibo
Introduction
  • With the rapid development of the Internet, information overload is becoming increasingly challenging in providing personalized information for users.
  • Some other methods [2, 10] consider that connected people would influence each other based on the social influence theory
  • These methods incorporate friends’ opinion on candidate items to model the user’s preference.
  • TrustSVD [10] incorporates the friends’ decisions to model the users’ preference and SAMN [2] design a two-level attention mechanism to model the friends’ influence
  • These methods only consider the first-order local neighbors’ information and neglect the helpful information from distant neighbors.
  • Neural Graph Collaborative Filtering (NGCF) [39] proposed an effective message propagation approach to aggregate the more similar information from neighbors, which enable the model to model the high-order relation information and filter the noisy neighbor effectively
Highlights
  • With the rapid development of the Internet, information overload is becoming increasingly challenging in providing personalized information for users
  • The performance of TrustSVD and TrustMF is better than BPR and FM, and SAMN, DiffNet++ and DICER outperform Neural Graph Collaborative Filtering (NGCF) and NCF
  • The substantial improvement of our model over the baselines could be attributed to two reasons: (1) our model use relationaware Graph Neural Networks (GNNs) to deal with the high-order social relation and collaborative similarity relation, which allow the related information from multi-relation neighbors to be utilized; (2) we model the users’ interests and items’ attraction based on the deep context, i.e., the graph enhanced user and item representation
  • We proposed DICER, which utilizes a multi-relation graph neural network to learn the graph information and extract the most related information under the graph enhanced deep context
  • Our method is equipped with good performance because: i) The multi-relation graph neural network module can capture the highorder relation information in both social graph and collaborative similarity graph. ii) The dual side deep context-aware modulation can model the rich user interest and item attraction from the interaction history
  • Our method DICER achieves the best performance on the two datasets
  • Our comparative experiments and ablation studies on the two benchmark datasets showed that the multi-relation graph neural network module could model the better high-order relation information
Methods
  • The proposed DICER was implemented with PyTorch and the authors use the Xavier initializer [9] to initialize the model parameters.
  • The authors use 80% as a training set to learn parameters, 10% as a validation set to tune hyper-parameters and 10% as a test set for the final performance Models Social Domain √S HS Item Domain √The author HIs.
Results
  • This is consistent with previous work [2, 44, 48], which indicates that social information is helpful to improve the recommendation performance.
  • The substantial improvement of the model over the baselines could be attributed to two reasons: (1) the model use relationaware GNN to deal with the high-order social relation and collaborative similarity relation, which allow the related information from multi-relation neighbors to be utilized; (2) the authors model the users’ interests and items’ attraction based on the deep context, i.e., the graph enhanced user and item representation
Conclusion
  • The authors proposed DICER, which utilizes a multi-relation graph neural network to learn the graph information and extract the most related information under the graph enhanced deep context.
  • The authors' method is equipped with good performance because: i) The multi-relation graph neural network module can capture the highorder relation information in both social graph and collaborative similarity graph.
  • Ii) The dual side deep context-aware modulation can model the rich user interest and item attraction from the interaction history.
  • The authors' comparative experiments and ablation studies on the two benchmark datasets showed that the multi-relation graph neural network module could model the better high-order relation information.
  • The deep context-aware modulation plays a crucial role in both the user and item side.
  • The authors seek to deploy the method on real-world recommender systems
Summary
  • Introduction:

    With the rapid development of the Internet, information overload is becoming increasingly challenging in providing personalized information for users.
  • Some other methods [2, 10] consider that connected people would influence each other based on the social influence theory
  • These methods incorporate friends’ opinion on candidate items to model the user’s preference.
  • TrustSVD [10] incorporates the friends’ decisions to model the users’ preference and SAMN [2] design a two-level attention mechanism to model the friends’ influence
  • These methods only consider the first-order local neighbors’ information and neglect the helpful information from distant neighbors.
  • Neural Graph Collaborative Filtering (NGCF) [39] proposed an effective message propagation approach to aggregate the more similar information from neighbors, which enable the model to model the high-order relation information and filter the noisy neighbor effectively
  • Methods:

    The proposed DICER was implemented with PyTorch and the authors use the Xavier initializer [9] to initialize the model parameters.
  • The authors use 80% as a training set to learn parameters, 10% as a validation set to tune hyper-parameters and 10% as a test set for the final performance Models Social Domain √S HS Item Domain √The author HIs.
  • Results:

    This is consistent with previous work [2, 44, 48], which indicates that social information is helpful to improve the recommendation performance.
  • The substantial improvement of the model over the baselines could be attributed to two reasons: (1) the model use relationaware GNN to deal with the high-order social relation and collaborative similarity relation, which allow the related information from multi-relation neighbors to be utilized; (2) the authors model the users’ interests and items’ attraction based on the deep context, i.e., the graph enhanced user and item representation
  • Conclusion:

    The authors proposed DICER, which utilizes a multi-relation graph neural network to learn the graph information and extract the most related information under the graph enhanced deep context.
  • The authors' method is equipped with good performance because: i) The multi-relation graph neural network module can capture the highorder relation information in both social graph and collaborative similarity graph.
  • Ii) The dual side deep context-aware modulation can model the rich user interest and item attraction from the interaction history.
  • The authors' comparative experiments and ablation studies on the two benchmark datasets showed that the multi-relation graph neural network module could model the better high-order relation information.
  • The deep context-aware modulation plays a crucial role in both the user and item side.
  • The authors seek to deploy the method on real-world recommender systems
Tables
  • Table1: Summary of notations
  • Table2: Statistics of the datasets
  • Table3: Comparison of the methods. For social domain, we use "S" represent the social information and "HS" represent the high-order social information. For item domain, we use "I" represent the interest information and "HI" represent the high-order interest information. For user interest and item attraction, we use "SC" denotes shallow context-aware and "DC" denotes deep context-aware. And we use "DL" denote deep learning based methods
  • Table4: Comparisons of different methods on Two datasets. Best baselines are underlined. The proposed method achieves best performances on all metrics which are in boldface. The last column “RI” indicates the relative improvement of DICER over the corresponding baseline on average
  • Table5: Effect of deep context and modulation on Ciao
  • Table6: Effect of GNN module on Ciao
  • Table7: Effect of rgnn layer numbers on Ciao
Download tables as Excel
Related work
  • In this section, we briefly review the related work about the social recommendation, graph neural network techniques employed for recommendation, and the context-aware recommendation.

    Social Recommendation. In recent years, there are lots of works exploiting user’s social relations for improving the recommender system [33, 34, 42, 45]. Most of them assume that users’ preference is similar to or influenced by their friends, which can be suggested by social theories such as social homophily [26] and social influence [25]. According to the assumptions above, social regularization has been proposed to restrain the user embedding learning process in the latent factor based models [13, 15, 23, 24]. And TrustMF [45] model is proposed to model the mutual influence between users by mapping users into two low-dimensional space: truster space and trustee space and factorize the social trust matrix. By treating the social neighbors’ opinion as the auxiliary implicit feedbacks of the targeted user, TrustSVD [10] is proposed to incorporate the social influence from social neighbors on top of SVD++ [19]. Moreover, some recent studies like [2, 6, 38] and [3, 8, 21] leverage deep neural network and transfer learning or adversarial learning approach respectively, to learn a more complex representation or model the shared knowledge between social domain and item domain. However, comparing with our models in this paper, the common limitations of existing studies are: i) they did not leverage the high-order social relation and collaborative relation among users; ii) they ignore the related information from interaction history based on the relation enhanced deep context.
Funding
  • This work is funded by NSFC 61976114 and NSFC 61936012
  • This work is also partially supported by the research funding from ZTE Corporation
Study subjects and analysis
times pseudo negative samples: 8
, collaborative similarity threshold = 0.1, the LeakyReLU slope is 0.2. In the training process, as there are many more unobserved items for each user, we randomly select 8 times pseudo negative samples for each user at each iteration. Since each iteration we change the pseudo negative samples, each unobserved item gives a weak signal

Reference
  • Joan Bruna, Wojciech Zaremba, Arthur Szlam, and Yann LeCun. 2014. Spectral Networks and Locally Connected Networks on Graphs. In ICLR 2014: International Conference on Learning Representations (ICLR) 2014.
    Google ScholarLocate open access versionFindings
  • Chong Chen, Min Zhang, Yiqun Liu, and Shaoping Ma. 2019. Social attentional memory network: Modeling aspect-and friend-level differences in recommendation. In Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining. 177–185.
    Google ScholarLocate open access versionFindings
  • Chong Chen, Min Zhang, Chenyang Wang, Weizhi Ma, Minming Li, Yiqun Liu, and Shaoping Ma. 2019. An Efficient Adaptive Transfer Neural Network for Social-aware Recommendation. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 225–234.
    Google ScholarLocate open access versionFindings
  • Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering. In Advances in Neural Information Processing Systems 29, D. D. Lee, M. Sugiyama, U. V. Luxburg, I. Guyon, and R. Garnett (Eds.). 3844–3852.
    Google ScholarLocate open access versionFindings
  • Tyler Derr, Yao Ma, and Jiliang Tang. 2018. Signed Graph Convolutional Networks. In 2018 IEEE International Conference on Data Mining (ICDM). 929–934.
    Google ScholarLocate open access versionFindings
  • Wenqi Fan, Qing Li, and Min Cheng. 2018. Deep Modeling of Social Relations for Recommendation.. In AAAI. 8075–8076.
    Google ScholarFindings
  • Wenqi Fan, Yao Ma, Qing Li, Yuan He, Eric Zhao, Jiliang Tang, and Dawei Yin. 2019. Graph Neural Networks for Social Recommendation. In The World Wide Web Conference on. 417–426.
    Google ScholarFindings
  • Wenqi Fan, Yao Ma, Dawei Yin, Jianping Wang, Jiliang Tang, and Qing Li. 2019. Deep social collaborative filtering. In Proceedings of the 13th ACM Conference on Recommender Systems. 305–313.
    Google ScholarLocate open access versionFindings
  • Xavier Glorot and Yoshua Bengio. 2010. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics. 249–256.
    Google ScholarLocate open access versionFindings
  • Guibing Guo, Jie Zhang, and Neil Yorke-Smith. 2015. Trustsvd: Collaborative filtering with both the explicit and implicit influence of user trust and of item ratings.. In Aaai, Vol. 15. 123–125.
    Google ScholarLocate open access versionFindings
  • Guibing Guo, Jie Zhang, and Neil Yorke-Smith. 2016. A Novel Recommendation Model Regularized with User Trust and Item Ratings. IEEE Transactions on Knowledge and Data Engineering 28, 7 (2016), 1607–1620.
    Google ScholarLocate open access versionFindings
  • Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural Collaborative Filtering. In WWW ’17 Proceedings of the 26th International Conference on World Wide Web. 173–182.
    Google ScholarLocate open access versionFindings
  • Guang-Neng Hu, Xin-Yu Dai, Yunya Song, Shu-Jian Huang, and Jia-Jun Chen. 2015. A synthetic approach for recommendation: combining ratings, social relations, and reviews. In Proceedings of the 24th International Conference on Artificial Intelligence. 1756–1762.
    Google ScholarLocate open access versionFindings
  • Yifan Hu, Y. Koren, and C. Volinsky. 2008. Collaborative Filtering for Implicit Feedback Datasets. In 2008 Eighth IEEE International Conference on Data Mining. 263–272.
    Google ScholarLocate open access versionFindings
  • Mohsen Jamali and Martin Ester. 2010. A matrix factorization technique with trust propagation for recommendation in social networks. In Proceedings of the fourth ACM conference on Recommender systems. 135–142.
    Google ScholarLocate open access versionFindings
  • Meng Jiang, Peng Cui, Fei Wang, Wenwu Zhu, and Shiqiang Yang. 2014. Scalable Recommendation with Social Contextual Information. IEEE Transactions on Knowledge and Data Engineering 26, 11 (2014), 2789–2802.
    Google ScholarLocate open access versionFindings
  • Diederik P. Kingma and Jimmy Lei Ba. 2015. Adam: A Method for Stochastic Optimization. In ICLR 2015: International Conference on Learning Representations 2015.
    Google ScholarLocate open access versionFindings
  • Thomas N. Kipf and Max Welling. 2017. Semi-Supervised Classification with Graph Convolutional Networks. In ICLR 2017: International Conference on Learning Representations 2017.
    Google ScholarLocate open access versionFindings
  • Yehuda Koren. 2008. Factorization meets the neighborhood: a multifaceted collaborative filtering model. In Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data mining. 426–434.
    Google ScholarLocate open access versionFindings
  • Y. Koren, R. Bell, and C. Volinsky. 2009. Matrix Factorization Techniques for Recommender Systems. IEEE Computer 42, 8 (2009), 30–37.
    Google ScholarLocate open access versionFindings
  • Adit Krishnan, Hari Cheruvu, Cheng Tao, and Hari Sundaram. 2019. A Modular Adversarial Approach to Social Recommendation. In Proceedings of the 28th ACM International Conference on Information and Knowledge Management. 1753–1762.
    Google ScholarLocate open access versionFindings
  • Wang Li. 2012. Context-Aware Recommender Systems. Journal of Software (2012).
    Google ScholarLocate open access versionFindings
  • Hao Ma, Haixuan Yang, Michael R. Lyu, and Irwin King. 2008. SoRec: Social Recommendation Using Probabilistic Matrix Factorization. In Proceedings of the 17th ACM Conference on Information and Knowledge Management (CIKM ’08). Association for Computing Machinery, New York, NY, USA, 931–940.
    Google ScholarLocate open access versionFindings
  • Hao Ma, Dengyong Zhou, Chao Liu, Michael R. Lyu, and Irwin King. 2011. Recommender systems with social regularization. In Proceedings of the fourth ACM international conference on Web search and data mining. 287–296.
    Google ScholarLocate open access versionFindings
  • Peter V. Marsden and Noah E. Friedkin. 1993. Network Studies of Social Influence. Sociological Methods & Research 22, 1 (1993), 127–151.
    Google ScholarLocate open access versionFindings
  • Miller McPherson, Lynn Smith-Lovin, and James M Cook. 2001. Birds of a Feather: Homophily in Social Networks. Review of Sociology 27, 1 (2001), 415–444.
    Google ScholarLocate open access versionFindings
  • Jiarui Qin, Kan Ren, Yuchen Fang, Weinan Zhang, and Yong Yu. 2020. Sequential Recommendation with Dual Side Neighbor-based Collaborative Relation Modeling. In WSDM 2020: The 13th ACM International Conference on Web Search and Data Mining. 465–473.
    Google ScholarLocate open access versionFindings
  • Zhaochun Ren, Shangsong Liang, Piji Li, Shuaiqiang Wang, and Maarten de Rijke. 2017. Social Collaborative Viewpoint Regression with Explainable Recommendations. In Proceedings of the Tenth ACM International Conference on Web Search and Data Mining. 485–494.
    Google ScholarLocate open access versionFindings
  • Steffen Rendle. 2010. Factorization machines. In 2010 IEEE International Conference on Data Mining. IEEE, 995–1000.
    Google ScholarLocate open access versionFindings
  • Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme. 2009. BPR: Bayesian personalized ranking from implicit feedback. In UAI ’09 Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence. 452–461.
    Google ScholarLocate open access versionFindings
  • Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl. 2001. Item-based collaborative filtering recommendation algorithms. In Proceedings of the 10th international conference on World Wide Web. 285–295.
    Google ScholarLocate open access versionFindings
  • Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research 15, 1 (2014), 1929–1958.
    Google ScholarLocate open access versionFindings
  • Jiliang Tang, Charu Aggarwal, and Huan Liu. 2016. Recommendations in Signed Social Networks. In WWW ’16 Proceedings of the 25th International Conference on World Wide Web. 31–40.
    Google ScholarLocate open access versionFindings
  • Jiliang Tang, Xia Hu, Huiji Gao, and Huan Liu. 2013. Exploiting local and global social context for recommendation. In IJCAI ’13 Proceedings of the Twenty-Third international joint conference on Artificial Intelligence. 2712–2718.
    Google ScholarLocate open access versionFindings
  • Jiliang Tang, Xia Hu, and Huan Liu. 2013. Social recommendation: a review. Social Network Analysis and Mining 3, 4 (2013), 1113–1133.
    Google ScholarFindings
  • Rianne van den Berg, Thomas N. Kipf, and Max Welling. 2017. Graph Convolutional Matrix Completion. arXiv preprint arXiv:1706.02263 (2017).
    Findings
  • Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. 2017. Graph attention networks. arXiv preprint arXiv:1710.10903 (2017).
    Findings
  • Xiang Wang, Xiangnan He, Liqiang Nie, and Tat-Seng Chua. 2017. Item Silk Road: Recommending Items from Information Domains to Social Users. In Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. 185–194.
    Google ScholarLocate open access versionFindings
  • Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 2019. Neural graph collaborative filtering. In Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval. 165–174.
    Google ScholarLocate open access versionFindings
  • Le Wu, Junwei Li, Peijie Sun, Richang Hong, Yong Ge, and Meng Wang. 2020. DiffNet++: A Neural Influence and Interest Diffusion Network for Social Recommendation. CoRR abs/2002.00844 (2020). arXiv:2002.00844 https://arxiv.org/abs/2002.00844
    Findings
  • Le Wu, Peijie Sun, Yanjie Fu, Richang Hong, Xiting Wang, and Meng Wang. 2019. A Neural Influence Diffusion Model for Social Recommendation. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 235–244.
    Google ScholarLocate open access versionFindings
  • Le Wu, Peijie Sun, Richang Hong, Yong Ge, and Meng Wang. 2021. Collaborative Neural Social Recommendation. IEEE Transactions on Systems, Man, and Cybernetics 51, 1 (2021), 464–476.
    Google ScholarLocate open access versionFindings
  • Qitian Wu, Hengrui Zhang, Xiaofeng Gao, Peng He, Paul Weng, Han Gao, and Guihai Chen. 2019. Dual graph attention networks for deep latent representation of multifaceted social effects in recommender systems. In The World Wide Web Conference. 2091–2102.
    Google ScholarFindings
  • Lin Xiao, Zhang Min, Zhang Yongfeng, Liu Yiqun, and Ma Shaoping. 2017. Learning and Transferring Social and Item Visibilities for Personalized Recommendation. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management. 337–346.
    Google ScholarLocate open access versionFindings
  • Bo Yang, Yu Lei, Jiming Liu, and Wenjie Li. 2017. Social Collaborative Filtering by Trust. IEEE Transactions on Pattern Analysis and Machine Intelligence 39, 8 (2017), 1633–1647.
    Google ScholarLocate open access versionFindings
  • Rex Ying, Ruining He, Kaifeng Chen, Pong Eksombatchai, William L. Hamilton, and Jure Leskovec. 2018. Graph Convolutional Neural Networks for Web-Scale Recommender Systems. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 974–983.
    Google ScholarLocate open access versionFindings
  • Wenhui Yu, Huidi Zhang, Xiangnan He, Xu Chen, Li Xiong, and Zheng Qin. 2018. Aesthetic-based Clothing Recommendation. In WWW 2018: International World Wide Web Conferences. 649–658.
    Google ScholarLocate open access versionFindings
  • Tong Zhao, Julian McAuley, and Irwin King. 2014. Leveraging Social Connections to Improve Personalized Ranking for Collaborative Filtering. In Proceedings of the
    Google ScholarLocate open access versionFindings
  • Guorui Zhou, Xiaoqiang Zhu, Chenru Song, Ying Fan, Han Zhu, Xiao Ma, Yanghui Yan, Junqi Jin, Han Li, and Kun Gai. 2018. Deep Interest Network for ClickThrough Rate Prediction. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 1059–1068.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments