Contrastive collaborative filtering method based on graph diffusion generation and adaptive sampling​

doi:10.11772/j.issn.1001-9081.2025060729

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (6): 1818-1828.DOI: 10.11772/j.issn.1001-9081.2025060729

• Data science and technology • Previous Articles

Contrastive collaborative filtering method based on graph diffusion generation and adaptive sampling

Hang QI, Tingting DONG, Yongqiang NAI, Xian MO()

School of Information Engineering，Ningxia University，Yinchuan Ningxia 750021，China

Received:2025-07-01 Revised:2025-10-02 Accepted:2025-10-14 Online:2025-10-30 Published:2026-06-10
Contact: Xian MO
About author:QI Hang， born in 2000， M. S. candidate. His research interests include graph learning， recommender systems.
DONG Tingting， born in 2000， M. S. candidate. Her research interests include graph learning， recommender systems.
NAI Yongqiang， born in 1985， Ph. D.， associate professor. His research interests include reinforcement learning， intelligent collaborative control， intelligent fault diagnosis， adaptive learning control.
First author contact:MO Xian， born in 1990， Ph. D.， associate professor. His research interests include graph learning， recommender systems.
Supported by:
National Natural Science Foundation of China(62306157);Natural Science Foundation of Ningxia(2024AAC05011)

基于图扩散生成与自适应采样的对比协同过滤方法

戚航, 董婷婷, 乃永强, 莫先()

宁夏大学信息工程学院，银川 750021

通讯作者: 莫先
作者简介:戚航（2000—），男，安徽淮南人，硕士研究生，CCF会员，主要研究方向：图学习、推荐系统
董婷婷（2000—），女，宁夏固原人，硕士研究生，主要研究方向：图学习、推荐系统
乃永强（1985—），男，甘肃庆阳人，副教授，博士，主要研究方向：强化学习、智能协同控制、智能故障诊断、自适应学习控制
第一联系人：莫先（1990—），男，四川南充人，副教授，博士，CCF会员，主要研究方向：图学习、推荐系统。
基金资助:
国家自然科学基金资助项目(62306157);国家自然科学基金资助项目(62263027);宁夏自然科学基金资助项目(2024AAC05011);宁夏自然科学基金资助项目(2024AAC05005)

Abstract

Abstract:

Aiming at the problems of the existing Graph Neural Network （GNN）-based collaborative filtering methods under sparse and noisy data conditions， such as the obscuring of true signals by static noise injection， the inability of fixed semantic prototypes to capture dynamic user interests， and the high computational overhead of complex augmentation， a graph diffusion generation and adaptive sampling-based contrastive collaborative filtering method was proposed. Firstly， a lightweight graph diffusion generation mechanism based on gradual denoising was designed， so as to optimize node representations through forward noise-adding and reverse denoising， thereby generating noise-resistant contrastive views. Then， random masking was integrated with Random Walk with Restart （RWR） to model local neighborhood features and global structural semantics collaboratively， thereby generating high-quality negative samples. Finally， an improved InfoNCE （Information Noise Contrastive Estimation） loss function was introduced to optimize the multi-view contrastive learning objective and enhance the discriminative power of representations. Experimental results on Gowalla， Yelp， and Amazon datasets show that compared to the best-performing baseline method， the proposed method improves the Top-20 Recall （Recall@20） by 0.63%， 1.36%， and 1.88%， respectively， and the Top-40 Normalized Discounted Cumulative Gain （NDCG@40） by 0.95%， 1.47%， and 1.24%， respectively， as well as improves the recommendation performance for long-tail users by 26.7%， increases the training efficiency by 90%， and accelerates the convergence speed by 32%. It can be seen that the proposed method enhances the noise resistance and dynamic adaptability of recommendation systems in open environments significantly.

Key words: Graph Neural Network (GNN), contrastive learning, self-supervised learning, diffusion process, graph algorithm

摘要：

针对现有基于图神经网络（GNN）的协同过滤方法在数据稀疏和噪声场景下存在的静态噪声注入易掩盖真实信号、固定语义原型难以捕捉用户的动态兴趣以及复杂增强方法计算开销大等问题，提出一种基于图扩散生成与自适应采样的对比协同过滤方法。首先，设计基于渐进去噪的轻量级图扩散生成机制，通过前向加噪和反向去噪优化节点表示，生成抗噪性强的对比视图；其次，结合随机掩码与重启随机游走算法（RWR），协同建模局部邻域特征与全局结构语义，生成高质量的负样本；最后，通过改进的InfoNCE （Information Noise Contrastive Estimation）损失函数优化多视图对比学习目标，并提升表征的判别性。在Gowalla、Yelp和Amazon数据集上，所提方法的前20个结果的召回率（Recall@20）指标较最优的基线方法分别提升了0.63%、1.36%和1.88%，前40个结果的归一化折损累计增益（NDCG@40）指标分别提升了0.95%、1.47%和1.24%，长尾用户的推荐效果提升了26.7%，训练效率提升了90%且收敛速度提升了32%。可见，该方法显著提升了开放环境下推荐系统的抗噪性与动态适应性。

关键词: 图神经网络, 对比学习, 自建督学习, 扩散过程, 图算法

CLC Number:

TP183

Hang QI, Tingting DONG, Yongqiang NAI, Xian MO. Contrastive collaborative filtering method based on graph diffusion generation and adaptive sampling[J]. Journal of Computer Applications, 2026, 46(6): 1818-1828.

戚航, 董婷婷, 乃永强, 莫先. 基于图扩散生成与自适应采样的对比协同过滤方法[J]. 《计算机应用》唯一官方网站, 2026, 46(6): 1818-1828.

Figures/Tables 10

Fig. 1 Method framework

Tab. 1 Statistics of datasets

数据集	用户数	项目数	交互数	交互密度/ $10 - 4$
Gowalla	25 557	19 747	294 983	5.85
Yelp	42 712	26 822	182 357	1.59
Amazon	76 469	83 761	966 680	1.51

Tab. 1 Statistics of datasets

数据集	用户数	项目数	交互数	交互密度/ $10 - 4$
Gowalla	25 557	19 747	294 983	5.85
Yelp	42 712	26 822	182 357	1.59
Amazon	76 469	83 761	966 680	1.51

Tab. 2 Performance comparison of different methods on three datasets

方法	Gowalla				Yelp				Amazon
方法	Recall@ 20	NDCG@20	Recall@40	NDCG@40	Recall@20	NDCG@20	Recall@40	NDCG@40	Recall@20	NDCG@20	Recall@40	NDCG@40
BiasMF	0.086 7	0.057 9	0.126 9	0.069 5	0.019 8	0.009 4	0.030 7	0.012 0	0.032 4	0.021 1	0.057 8	0.029 3
NCF	0.101 9	0.067 4	0.156 3	0.083 3	0.030 4	0.014 3	0.048 7	0.018 7	0.036 7	0.023 4	0.060 0	0.030 6
AutoRec	0.147 7	0.069 0	0.251 1	0.098 5	0.049 1	0.022 2	0.069 2	0.026 8	0.052 5	0.031 8	0.082 6	0.041 5
PinSage	0.123 5	0.080 9	0.188 2	0.099 4	0.051 0	0.024 5	0.074 3	0.031 5	0.048 6	0.031 7	0.077 3	0.040 2
STGCN	0.157 4	0.104 2	0.231 8	0.125 2	0.056 2	0.028 2	0.085 6	0.035 5	0.058 3	0.037 7	0.090 8	0.047 8
GCMC	0.186 3	0.115 1	0.262 7	0.139 0	0.058 4	0.028 0	0.089 1	0.036 0	0.083 7	0.057 9	0.119 6	0.069 2
NGCF	0.175 7	0.113 5	0.258 6	0.136 7	0.068 1	0.033 6	0.101 9	0.041 9	0.055 1	0.035 3	0.087 6	0.045 4
GCCF	0.201 2	0.128 2	0.290 3	0.153 2	0.074 2	0.036 5	0.115 1	0.046 6	0.077 2	0.050 1	0.117 5	0.062 5
LightGCN	0.223 0	0.143 3	0.318 1	0.167 0	0.076 1	0.037 3	0.117 5	0.047 4	0.086 8	0.057 1	0.128 5	0.069 7
DGCF	0.205 5	0.131 2	0.292 9	0.155 5	0.070 0	0.034 7	0.107 2	0.043 7	0.061 7	0.037 2	0.091 2	0.046 8
SLRec	0.200 1	0.129 8	0.286 3	0.154 0	0.066 5	0.032 7	0.103 2	0.041 8	0.074 2	0.048 0	0.112 3	0.059 8
NCL	0.228 3	0.147 8	0.323 2	0.174 5	0.080 6	0.040 2	0.123 0	0.050 5	0.095 5	0.062 3	0.140 9	0.076 4
SGL	0.233 2	0.150 9	0.325 1	0.178 0	0.080 3	0.039 8	0.122 6	0.050 2	0.087 4	0.569 0	0.131 2	0.070 4
HCCF	0.229 3	0.148 2	0.325 8	0.175 1	0.078 9	0.039 1	0.121 0	0.049 2	0.088 5	0.057 8	0.133 5	0.071 6
AdaGCL	0.133 5	0.082 7	0.192 8	0.099 0	0.086 3	0.043 4	0.131 6	0.054 5	0.060 1	0.046 9	0.094 0	0.057 6
AutoCF	0.253 8	0.164 5	0.344 1	0.189 8	0.086 9	0.043 7	0.127 3	0.053 3	0.127 7	0.087 9	0.178 2	0.104 8
MNGAE	0.247 3	0.154 1	0.314 5	0.190 4	0.088 1	0.044 1	0.122 6	0.044 3	0.103 3	0.068 8	0.166 7	0.094 3
本文方法	0.255 4	0.165 6	0.352 2	0.192 2	0.089 3	0.050 1	0.133 9	0.055 3	0.130 1	0.091 0	0.180 3	0.106 1

Fig. 2 Performance comparison of different variants

Fig. 3 Impact analysis comparison of different hyperparameters on three datasets

Fig. 4 Visualization of Transformer attention weights

Fig. 5 Dynamic visualization of diffusion process?

Fig. 6 Visualization of RWR subgraph mask effects?

Fig. 7 Performance comparison for different user types

Fig. 8 Experimental results of convergence analysis

References 39

[1]	WU L， HE X， WANG X， et al. A survey on accuracy-oriented neural recommendation： from collaborative filtering to information-rich recommendation［J］. IEEE Transactions on Knowledge and Data Engineering， 2023， 35（5）： 4425-4445.
[2]	KOREN， YBELL R， VOLINSKY C. Matrix factorization techniques for recommender systems［J］. Computer， 2009， 42（8）： 30-37.
[3]	SEDHAIN S， MENON A K， SANNER S， et al. AutoRec： autoencoders meet collaborative filtering［C］// Proceedings of the 24th International Conference on World Wide Web. New York： ACM， 2015： 111-112.
[4]	CHEN J， ZHANG H， HE X， et al. Attentive collaborative filtering： multimedia recommendation with item-and component-level attention［C］// Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2017： 335-344.
[5]	VAN DEN BERG R， KIPF T N， WELLING M. Graph convolutional matrix completion［EB/OL］. ［2025-06-28］..
[6]	WANG X， HE X， WANG M， et al. Neural graph collaborative filtering［C］// Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2019： 165-174.
[7]	HE X， DENG K， WANG X， et al. LightGCN： simplifying and powering graph convolution network for recommendation［C］// Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2020： 639-648.
[8]	CHEN L， WU L， HONG R， et al. Revisiting graph based collaborative filtering： a linear residual graph convolutional network approach［C］// Proceedings of the 34th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2020： 27-34.
[9]	WU J， WANG X， FENG F， et al. Self-supervised graph learning for recommendation［C］// Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2021： 726-735.
[10]	LIU Z， MA Y， OUYANG Y， et al. Contrastive learning for recommender system［EB/OL］. ［2025-06-24］..
[11]	ZHANG S， CHEN L， WANG C， et al. Temporal graph contrastive learning for sequential recommendation［C］// Proceedings of the 38th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2024： 9359-9367.
[12]	YAO T， YI X， CHENG D Z， et al. Self-supervised learning for large-scale item recommendations［C］// Proceedings of the 30th ACM International Conference on Information and Knowledge Management. New York： ACM， 2021： 4321-4330.
[13]	YANG Y， HUANG C， XIA L， et al. Debiased contrastive learning for sequential recommendation［C］// Proceedings of the ACM Web Conference 2023. New York： ACM， 2023： 1063-1073.
[14]	LI Y， CHEN H， TAN J， et al. Causal factorization machine for robust recommendation［C］// Proceedings of the 22nd ACM/IEEE Joint Conference on Digital Libraries. New York： ACM， 2022： 1-9.
[15]	XIA L， HUANG C， XU Y， et al. Hypergraph contrastive collaborative filtering［C］// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2022： 70-79.
[16]	LIN Z， TIAN C， HOU Y， et al. Improving graph collaborative filtering with neighborhood-enriched contrastive learning［C］// Proceedings of the ACM Web Conference 2022. New York： ACM， 2022： 2320-2329.
[17]	DU Y， ZHU X， CHEN L， et al. MetaKG： meta-learning on knowledge graph for cold-start recommendation［J］. IEEE Transactions on Knowledge and Data Engineering， 2023， 35（10）： 9850-9863.
[18]	WU M， ZHENG X， ZHANG Q， et al. Graph learning under distribution shifts： a comprehensive survey on domain adaptation， out-of-distribution， and continual learning［EB/OL］. ［2025-06-28］..
[19]	TANG J， DAI S， SUN Z， et al. Towards robust recommendation via decision boundary-aware graph contrastive learning［C］// Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2024： 2854-2865.
[20]	LIU C， YU C， GUI N， et al. SimGCL： graph contrastive learning by finding homophily in heterophily［J］. Knowledge and Information Systems， 2024， 66（3）： 2089-2114.
[21]	HO J， JAIN A， ABBEEL P. Denoising diffusion probabilistic models［C］// Proceedings of the 34th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2020： 6840-6851.
[22]	GOODFELLOW I， POUGET-ABADIE J， MIRZA M， et al. Generative adversarial networks［J］. Communications of the ACM， 2020， 63（11）： 139-144.
[23]	CAI X， HUANG C， XIA L， et al. LightGCL： simple yet effective graph contrastive learning for recommendation［EB/OL］. ［2025-06-28］..
[24]	LI Z， SUN A， LI C. DiffuRec： a diffusion model for sequential recommendation［J］. ACM Transactions on Information Systems， 2024， 42（3）： No.66.
[25]	GASTEIGER J， WEIßENBERGER S， GÜNNEMANN S. Diffusion improves graph learning［EB/OL］. ［2025-06-28］. .
[26]	RUSAK E， REIZINGER P， JUHOS A， et al. InfoNCE： identifying the gap between theory and practice［C］// Proceedings of the 28th International Conference on Artificial Intelligence and Statistics. New York： JMLR.org， 2025： 4159-4167.
[27]	PEROZZI B， AI-RFOU R， SKIENA S. DeepWalk： online learning of social representations［C］// Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2014： 701-710.
[28]	GROVER A， LESKOVEC J. node2vec： scalable feature learning for networks［C］// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2016： 855-864.
[29]	ROMBACH R， BLATTMANN A， LORENZ D， et al. High-resolution image synthesis with latent diffusion models［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 10674-10685.
[30]	ZHANG S， TONG H， XU J， et al. Graph convolutional networks： a comprehensive review［J］. Computational Social Networks， 2019， 6： No.11.
[31]	CHO E， MYERS S A， LESKOVEC J. Friendship and mobility： user movement in location-based social networks ［C］// Proceedings of the 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2011： 1082-1090.
[32]	ASGHAR N. Yelp dataset challenge： review rating prediction ［EB/OL］. ［2025-06-01］. .
[33]	HE X， LIAO L， ZHANG H， et al. Neural collaborative filtering ［C］// Proceedings of the 26th International Conference on World Wide Web. New York： ACM， 2017： 173-182.
[34]	YING R， HE R， CHEN K， et al. Graph convolutional neural networks for web-scale recommender systems［C］// Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2018： 974-983.
[35]	YAN S， XIONG Y， LIN D. Spatial temporal graph convolutional networks for skeleton-based action recognition ［EB/OL］. ［2025-06-02］. .
[36]	WANG X， JIN H， ZHANG A， et al. Disentangled graph collaborative filtering［C］// Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2020： 1001-1010.
[37]	赖庆涛，刘淇，朱琛，等. 融合掩码和邻域增强的图协同过滤算法［J］. 软件学报， 2023， 34（8）： 3780-3794.
	LAI Q T， LIU Q， ZHU C， et al. A graph collaborative filtering algorithm integrating mask and neighborhood enhancement［J］. Journal of Software， 2023， 34（8）： 3780-3794.
[38]	JIANG Y， HUANG C， HUANG L. Adaptive graph contrastive learning for recommendation［C］// Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2023： 4252-4261.
[39]	XIA L， HUANG C， HUANG C， et al. Automated self-supervised learning for recommendation［C］// Proceedings of the ACM Web Conference 2023. New York： ACM， 2023： 992-1002.