Full Text:  <1137>

Summary:  <220>

CLC number: TP391

On-line Access: 2023-01-21

Received: 2022-04-07

Revision Accepted: 2023-01-21

Crosschecked: 2022-08-02

Cited: 0

Clicked: 1165

Citations:  Bibtex RefMan EndNote GB/T7714

 ORCID:

Yitong WANG

https://orcid.org/0000-0003-4020-5767

Fei CAI

https://orcid.org/0000-0002-5709-1682

Chengyu SONG

https://orcid.org/0000-0001-7020-7603

-   Go to

Article info.
Open peer comments

Frontiers of Information Technology & Electronic Engineering 

Accepted manuscript available online (unedited version)


Self-supervised graph learning with target-adaptive masking for session-based recommendation


Author(s):  Yitong WANG, Fei CAI, Zhiqiang PAN, Chengyu SONG

Affiliation(s):  Science and Technology on Information Systems Engineering Laboratory, National University of Defense Technology, Changsha 410073, China

Corresponding email(s):  wangyitong20@nudt.edu.cn, caifei08@nudt.edu.cn, panzhiqiang@nudt.edu.cn, songchengyu@nudt.edu.cn

Key Words:  Session-based recommendation; Self-supervised learning; Graph neural networks; Target-adaptive masking


Share this article to: More <<< Previous Paper|Next Paper >>>

Yitong WANG, Fei CAI, Zhiqiang PAN, Chengyu SONG. Self-supervised graph learning with target-adaptive masking for session-based recommendation[J]. Frontiers of Information Technology & Electronic Engineering,in press.https://doi.org/10.1631/FITEE.2200137

@article{title="Self-supervised graph learning with target-adaptive masking for session-based recommendation",
author="Yitong WANG, Fei CAI, Zhiqiang PAN, Chengyu SONG",
journal="Frontiers of Information Technology & Electronic Engineering",
year="in press",
publisher="Zhejiang University Press & Springer",
doi="https://doi.org/10.1631/FITEE.2200137"
}

%0 Journal Article
%T Self-supervised graph learning with target-adaptive masking for session-based recommendation
%A Yitong WANG
%A Fei CAI
%A Zhiqiang PAN
%A Chengyu SONG
%J Frontiers of Information Technology & Electronic Engineering
%P 73-87
%@ 2095-9184
%D in press
%I Zhejiang University Press & Springer
doi="https://doi.org/10.1631/FITEE.2200137"

TY - JOUR
T1 - Self-supervised graph learning with target-adaptive masking for session-based recommendation
A1 - Yitong WANG
A1 - Fei CAI
A1 - Zhiqiang PAN
A1 - Chengyu SONG
J0 - Frontiers of Information Technology & Electronic Engineering
SP - 73
EP - 87
%@ 2095-9184
Y1 - in press
PB - Zhejiang University Press & Springer
ER -
doi="https://doi.org/10.1631/FITEE.2200137"


Abstract: 
Session-based recommendation aims to predict the next item based on a user's limited interactions within a short period. Existing approaches use mainly recurrent neural networks (RNNs) or graph neural networks (GNNs) to model the sequential patterns or the transition relationships between items. However, such models either ignore the over-smoothing issue of GNNs, or directly use cross-entropy loss with a softmax layer for model optimization, which easily results in the over-fitting problem. To tackle the above issues, we propose a self-supervised graph learning with target-adaptive masking (SGL-TM) method. Specifically, we first construct a global graph based on all involved sessions and subsequently capture the self-supervised signals from the global connections between items, which helps supervise the model in generating accurate representations of items in the ongoing session. After that, we calculate the main supervised loss by comparing the ground truth with the predicted scores of items adjusted by our designed target-adaptive masking module. Finally, we combine the main supervised component with the auxiliary self-supervision module to obtain the final loss for optimizing the model parameters. Extensive experimental results from two benchmark datasets, Gowalla and Diginetica, indicate that SGL-TM can outperform state-of-the-art baselines in terms of Recall@20 and MRR@20, especially in short sessions.

融合自监督图学习与目标自适应屏蔽的会话型推荐方法

王祎童,蔡飞,潘志强,宋城宇
国防科技大学信息系统工程重点实验室,中国长沙市,410073
摘要:会话型推荐旨在根据用户在短时间内有限的交互来预测下一个时间戳将要进行交互的物品。现有模型主要使用循环神经网络(RNN)或图神经网络(GNN)来建模顺序序列或物品之间的传递关系。然而,此类模型要么忽略了GNN的过度平滑问题,要么直接利用交叉熵损失和softmax层进行模型优化,容易导致过拟合问题。为了解决上述问题,本文提出一种融合自监督图学习与目标自适应屏蔽的会话型推荐方法(SGL-TM)。具体来说,首先根据所有涉及到的会话构建全局图,然后从物品之间的全局连接中捕捉自监督信号,用来监督模型生成当前会话中准确的物品表示。之后,通过比较真值与经过我们设计的目标自适应屏蔽模块调整后的物品的预测分数来计算主监督损失。最后,将主监督组件与辅助自监督模块相结合,以获得用来优化模型参数的最终损失。在两个真实数据集(Gowalla和Diginetica)上的大量实验结果表明,SGL-TM在Recall@20和MRR@20方面的性能优于最先进的基准模型,尤其体现在短会话上。

关键词组:会话型推荐;自监督学习;图神经网络;目标自适应屏蔽

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference

[1]Afouras T, Owens A, Chung JS, et al., 2020. Self-supervised learning of audio-visual objects from video. Proc 16th European Conf, p.208-224.

[2]Chen M, Wei ZW, Huang ZF, et al., 2020. Simple and deep graph convolutional networks. Proc 37th Int Conf on Machine Learning, p.1725-1735.

[3]Chen T, Kornblith S, Norouzi M, et al., 2020. A simple framework for contrastive learning of visual representations. Proc 37th Int Conf on Machine Learning, p.1597-1607.

[4]Chen TW, Wong RCW, 2020. Handling information loss of graph neural networks for session-based recommendation. Proc 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining, p.1172-1180.

[5]Cheng ZY, Shen JL, Zhu L, et al., 2017. Exploiting music play sequence for music recommendation. Proc 26th Int Joint Conf on Artificial Intelligence, p.3654-3660.

[6]Choi M, Kim J, Lee J, et al., 2021. Session-aware linear item–item models for session-based recommendation. Proc Web Conf, p.2186-2197.

[7]Davidson J, Liebald B, Liu JN, et al., 2010. The YouTube video recommendation system. Proc 4th ACM Conf on Recommender Systems, p.293-296.

[8]Hidasi B, Karatzoglou A, Baltrunas L, et al., 2016. Session-based recommendations with recurrent neural networks. Proc 4th Int Conf on Learning Representations.

[9]Hjelm RD, Fedorov A, Lavoie-Marchildon S, et al., 2019. Learning deep representations by mutual information estimation and maximization. Proc 7th Int Conf on Learning Representations.

[10]Kingma DP, Ba LJ, 2015. Adam: a method for stochastic optimization. Proc 3rd Int Conf on Learning Representations.

[11]Kipf TN, Welling M, 2017. Semi-supervised classification with graph convolutional networks. Proc 5th Int Conf on Learning Representations.

[12]Kong LP, de Masson d’Autume C, Yu L, et al., 2020. A mutual information maximization perspective of language representation learning. Proc 8th Int Conf on Learning Representations.

[13]Li J, Ren PJ, Chen ZM, et al., 2017. Neural attentive session-based recommendation. Proc ACM Conf on Information and Knowledge Management, p.1419-1428.

[14]Li YJ, Tarlow D, Brockschmidt M, et al., 2016. Gated graph sequence neural networks. Proc 4th Int Conf on Learning Representations.

[15]Liu Q, Zeng YF, Mokhosi R, et al., 2018. STAMP: short-term attention/memory priority model for session-based recommendation. Proc 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining, p.1831-1839.

[16]Nie LQ, Li YQ, Feng FL, et al., 2020. Large-scale question tagging via joint question-topic embedding learning. ACM Trans Inform Syst, 38(2):20.

[17]Nie LQ, Jiao FK, Wang WJ, et al., 2021. Conversational image search. IEEE Trans Image Process, 30:7732-7743.

[18]Pan ZQ, Cai F, Chen WY, et al., 2020. Star graph neural networks for session-based recommendation. Proc 29th ACM Int Conf on Information & Knowledge Management, p.1195-1204.

[19]Pan ZQ, Cai F, Chen WY, et al., 2022. Graph co-attentive session-based recommendation. ACM Trans Inform Syst, 40(4):67.

[20]Qiu RH, Li JJ, Huang Z, et al., 2019. Rethinking the item order in session-based recommendation with graph neural networks. Proc 28th ACM Int Conf on Information and Knowledge Management, p.579-588.

[21]Qiu RH, Huang Z, Li JJ, et al., 2020. Exploiting cross-session information for session-based recommendation with graph neural networks. ACM Trans Inform Syst, 38(3):22.

[22]Rendle S, Freudenthaler C, Schmidt-Thieme L, 2010. Factorizing personalized Markov chains for next-basket recommendation. Proc 19th Int Conf on World Wide Web, p.811-820.

[23]Rumelhart DE, Hinton GE, Williams RJ, 1986. Learning representations by back-propagating errors. Nature, 323(6088):533-536.

[24]Shani G, Heckerman D, Brafman RI, 2005. An MDP-based recommender system. J Mach Learn Res, 6:1265-1295.

[25]Singhal A, Sinha P, Pant R, 2017. Use of deep learning in modern recommendation system: a summary of recent works. Int J Comput Appl, 180(7):17-22.

[26]Tan YK, Xu XX, Liu Y, 2016. Improved recurrent neural networks for session-based recommendations. Proc 1st Workshop on Deep Learning for Recommender Systems, p.17-22.

[27]Tang JX, Wang K, 2018. Personalized top-n sequential recommendation via convolutional sequence embedding. Proc 11th ACM Int Conf on Web Search and Data Mining, p.565-573.

[28]Veličković P, Cucurull G, Casanova A, et al., 2018. Graph attention networks. Proc 6th Int Conf on Learning Representations.

[29]Wang SJ, Cao LB, Wang Y, et al., 2022. A survey on session-based recommender systems. ACM Comput Surv, 54(7):154.

[30]Wang X, He XN, Wang M, et al., 2019. Neural graph collaborative filtering. Proc 42nd Int ACM SIGIR Conf on Research and Development in Information Retrieval, p.165-174.

[31]Wang XN, Tan QM, 2020. DAN: a deep association neural network approach for personalization recommendation. Front Inform Technol Electron Eng, 21(7):963-980.

[32]Wang ZY, Wei W, Cong G, et al., 2020. Global context enhanced graph neural networks for session-based recommendation. Proc 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval, p.169-178.

[33]Wu S, Tang YY, Zhu YQ, et al., 2019. Session-based recommendation with graph neural networks. Proc 33rd AAAI Conf on Artificial Intelligence, p.346-353.

[34]Xia X, Yin HZ, Yu JL, et al., 2021a. Self-supervised graph co-training for session-based recommendation. Proc 30th ACM Int Conf on Information & Knowledge Management, p.2180-2190.

[35]Xia X, Yin HZ, Yu JL, et al., 2021b. Self-supervised hyper- graph convolutional networks for session-based recommendation. Proc 35th AAAI Conf on Artificial Intelligence, p.4503-4511.

[36]Xie X, Sun F, Liu ZY, et al., 2022. Contrastive learning for sequential recommendation. Proc IEEE 38th Int Conf on Data Engineering, p.1259-1273.

[37]Xin X, Karatzoglou A, Arapakis I, et al., 2020. Self-supervised reinforcement learning for recommender systems. Proc 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval, p.931-940.

[38]Xu CF, Zhao PP, Liu YC, et al., 2019. Graph contextualized self-attention network for session-based recommendation. Proc 28th Int Joint Conf on Artificial Intelligence, p.3940-3946.

[39]Yao TS, Yi XY, Cheng DZ, et al., 2021. Self-supervised learning for large-scale item recommendations. Proc 30th ACM Int Conf on Information & Knowledge Management, p.4321-4330.

[40]Yuan FJ, Karatzoglou A, Arapakis I, et al., 2019. A simple convolutional generative network for next item recommendation. Proc 12th ACM Int Conf on Web Search and Data Mining, p.582-590.

[41]Zhang JQ, Zhao Y, Saleh M, et al., 2020. PEGASUS: pre-training with extracted gap-sentences for abstractive summarization. Proc 37th Int Conf on Machine Learning, p.11328-11339.

[42]Zheng JW, Ma QL, Gu H, et al., 2021. Multi-view denoising graph auto-encoders on heterogeneous information networks for cold-start recommendation. Proc 27th ACM SIGKDD Conf on Knowledge Discovery & Data Mining, p.2338-2348.

[43]Zhou K, Wang H, Zhao WX, et al., 2020. S3-Rec: self-supervised learning for sequential recommendation with mutual information maximization. Proc 29th ACM Int Conf on Information & Knowledge Management, p.1893-1902.

Open peer comments: Debate/Discuss/Question/Opinion

<1>

Please provide your name, email address and a comment





Journal of Zhejiang University-SCIENCE, 38 Zheda Road, Hangzhou 310027, China
Tel: +86-571-87952783; E-mail: cjzhang@zju.edu.cn
Copyright © 2000 - 2024 Journal of Zhejiang University-SCIENCE