سیستم توصیه‌گر اجتماعی مقاوم مبتنی بر یادگیری تقابلی گراف با نویز تخاصمی و ترازسازی دو - دامنه‌ای

مقالات آماده انتشار

نوع مقاله : مقاله پژوهشی

نویسندگان

1 استادیار گروه علوم کامپیوتر، دانشگاه سیستان و بلوچستان، زاهدان، ایران

2 استادیار گروه علوم کامپیوتر، دانشکده ریاضی، آمار و علوم کامپیوتر، دانشکدگان علوم، دانشگاه تهران، تهران، ایران

چکیده

بهره‌گیری از شبکه‌های اجتماعی به‌عنوان یک منبع اطلاعاتی کمکی، راهکاری مؤثر برای غلبه بر چالش پراکندگی داده‌ها در سیستم‌های توصیه‌گر محسوب می‌شود. بااین‌حال، روابط اجتماعی در دنیای واقعی غالباً نویزدار بوده و فرض «همسویی کامل ترجیحات میان دوستان» همواره صادق نیست. اخیراً، یادگیری تقابلی گراف برای مقابله با این نویزها موردتوجه قرارگرفته است، اما رویکردهای رایج با تکیه‌بر روش‌های تصادفی مانند حذف یال‌ها، ساختار معنایی گراف را مخدوش کرده و بازنمایی‌های ناپایداری تولید می‌کنند. برای رفع این محدودیت‌ها، در این مقاله یک چارچوب نوین مبتنی بر یادگیری تقابلی گراف با استفاده از نویز تخاصمی پیشنهادشده است. در این معماری، به‌جای ایجاد اختلال در ساختار گراف، با بهره‌گیری از روش گرادیان سریع، اختلالات هدفمندی به فضای پیوسته بازنمایی‌ها تزریق می‌شود تا نماهای تقابلی "سخت" و مقاومی تولید گردد. علاوه بر این، یک رویکرد یادگیری تقابلی دوگانه توسعه‌یافته است که شامل ترازسازی درون دامنه‌ای (برای استخراج ویژگی‌های خالص هر گراف) و میان‌دامنه‌ای (برای انتقال هدفمند سیگنال‌های مفید اجتماعی به دامنه تعاملی) است. ارزیابی‌های تجربی بر روی چهار مجموعه داده واقعی (Yelp, Epinions, Trustfilm, Ciao) نشان می‌دهد که مدل پیشنهادی در مقایسه با جدیدترین روش‌های پایه، به‌طور میانگین به میزان 5/8% در شاخص Recall و 2/7% در شاخص NDCG بهبود عملکرد داشته است. این نتایج اثربخشی استفاده از نویز تخاصمی را در استخراج بازنمایی‌های مقاوم و بهبود دقت توصیه‌ها تأیید می‌کند.

کلیدواژه‌ها

موضوعات

عنوان مقاله [English]

Robust Social Recommendation via Adversarial Graph Contrastive Learning and Dual-Domain Alignment

نویسندگان [English]

  • Mohammad Mehdi Keikha 1
  • Abolfazl Nadi 2
1 Assistant Professor, Department of Computer Science, University of Sistan and Baluchestan, Zahedan, Iran
2 Assistant Professor, Department of Computer Science, School of Mathematics, Statistics and Computer Science, College of Science, University of Tehran, Tehran, Iran
چکیده [English]

Leveraging social networks as an auxiliary information source is considered an effective solution to overcome the data sparsity challenge in recommender systems. However, real-world social relations are often noisy, and the assumption of "perfect preference alignment among friends" does not always hold true. Recently, Graph Contrastive Learning (GCL) has garnered attention to tackle this noise; nevertheless, conventional approaches relying on stochastic methods such as edge dropping distort the graph's semantic structure and yield unstable representations. To address these limitations, this paper proposes a novel framework based on graph contrastive learning utilizing adversarial noise. In this architecture, instead of perturbing the graph structure, targeted perturbations are injected into the continuous representation space using the Fast Gradient Method (FGM) to generate "hard" and robust contrastive views. Furthermore, a dual contrastive learning approach is developed, comprising intra-domain alignment (to extract the pure features of each graph) and cross-domain alignment (to purposively transfer useful social signals to the interaction domain). Empirical evaluations on four real-world datasets (Yelp, Epinions, Trustfilm, Ciao) demonstrate that the proposed model outperforms state-of-the-art baseline methods, achieving an average performance improvement of $8.5\%$ in the Recall metric and $7.2\%$ in the NDCG metric. These results verify the effectiveness of employing adversarial noise in extracting robust representations and improving recommendation accuracy.

کلیدواژه‌ها [English]

  • Social Recommender Systems
  • Graph Contrastive Learning
  • Adversarial Noise
  • Graph Neural Networks
  • Collaborative Filtering

مراجع

[1]     M. A. Shu-Ting, G. I. Felicia, L. Vivian, and C. H. Wang, “A survey of social recommender systems,” Social Netw. Anal. Mining, vol. 6, no. 1, pp. 1–21, 2016.
[2]     J. S. Breese, D. Heckerman, and C. Kadie, “Empirical analysis of predictive algorithms for collaborative filtering,” in Proc. 14th Conf. Uncertainty in Artif. Intell. (UAI), 1998, pp. 43–52.
[3]     X. Cai, C. Huang, L. Xia, and X. Ren, “LightGCL: Simple yet effective graph contrastive learning for recommendation,” in Proc. 11th Int. Conf. Learn. Represent. (ICLR), 2023.
[4]     S. Raza, M. Rahman, S. Kamawal, A. Toroghi, A. Raval, F. Navah, and A. Kazemeini, “A comprehensive review of recommender systems: Transitioning from theory to practice,” Comput. Sci. Rev., vol. 59, p. 100849, 2026.
[5]     F. Wu, A. H. Souri, L. J. Li, Y. Liu, and T. Mei, “A comprehensive survey on graph neural networks,” IEEE Trans. Neural Netw. Learn. Syst., vol. 32, no. 1, pp. 4–24, 2021.
[6]     M. M. Keikha, S. Barahoie, and A. Nadi, “Grace: A scalable framework for graph embedding with autoencoder-based feature compression and random walks,” Int. J. Data Sci. Anal., vol. 22, Art. no. 128, 2026, doi: 10.1007/s41060-026-01091-z.
[7]     C. Song, T. Wu, Y. Liu, J. Ge, and P. S. Yu, “Social recommendation with bi-polarized preference and inter-domain knowledge transfer,” in Proc. 28th ACM Int. Conf. Inf. Knowl. Manage. (CIKM), 2019, pp. 117–126.
[8]     M. M. Keikha and H. Rezaei, “CF-X: A multi-view chaotic-fuzzy network representation learning framework for heterophilic graphs,” Appl. Soft Comput., vol. 197, Art. no. 115127, 2026.
[9]     Y. Koren, R. Bell, and C. Volinsky, “Matrix factorization techniques for recommender systems,” Computer, vol. 42, no. 8, pp. 30–37, 2009.
[10] X. He, L. Liao, H. Zhang, L. Nie, X. Hu, and T. S. Chua, “Neural collaborative filtering,” in Proc. 26th Int. Conf. World Wide Web (WWW), 2017, pp. 173–182.
[11] M. M. Keikha and S. Barahouei, “A novel deep learning-based approach for graph dimensionality reduction by using fuzzy logic and random walks,” Applied and basic Machine intelligence research, vol. 2, no. 1, pp. 131–141, 2025 [In Persian].
[12] X. He, K. Deng, X. Wang, Y. Li, Y. Zhang, and M. Wang, “LightGCN: Simplifying and powering graph convolution network for recommendation,” in Proc. 43rd Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, 2020, pp. 639–648.
[13] J. Zhang, X. Lin, X. Wang, and Q. Li, “Adaptive diffusion in graph neural networks for recommendation,” Knowl.-Based Syst., vol. 230, Art. no. 107380, 2021.
[14] H. Ma, H. Yang, M. R. Lyu, and I. King, “SoRec: Social recommendation using probabilistic matrix factorization,” in Proc. 17th ACM Conf. Inf. Knowl. Manage. (CIKM), 2008, pp. 931–940.
[15] S. Jamali and M. Ester, “A matrix factorization technique with trust propagation for recommendation in social networks,” in Proc. 4th ACM Conf. Recommender Syst. (RecSys), 2010, pp. 135–142.
[16] W. Fan et al., “Graph neural networks for social recommendation,” in Proc. World Wide Web Conf. (WWW), 2019, pp. 417–426.
[17] L. Wu, J. Li, P. Cui, C. Li, B. Wang, and X. Wu, “Diffnet: A differential influence network for social recommendation,” IEEE Trans. Knowl. Data Eng., vol. 33, no. 4, pp. 1438–1449, 2021.
[18] Y. Liu, Q. Ren, S. Fu, and Y. Liu, “KAN-infused social recommendation: A contrastive graph learning approach with bidirectional feature fusion,” Inf. Fusion, vol. 125, Art. no. 103448, 2026.
[19] J. Wu, X. Wang, F. Feng, X. He, L. Chen, J. Lian, and X. Xie, “Self-supervised graph learning for recommendation,” in Proc. 44th Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, 2021, pp. 726–735.
[20] B. Wu, B. Zhang, Y. Tian, C. Li, J. Liang, and Y. Ye, “EGCL: An effective and efficient graph contrastive learning framework for social recommendation,” ACM Trans. Inf. Syst., vol. 44, no. 3, Art. no. 58, Feb. 2026.
[21] T. Yang, T. Chen, X. Wang, and Z. Guan, “RCGRL: Reinforcement-optimized contrastive graph representation learning for social recommendation,” arXiv preprint, 2024.
[22] Y. Xie, J. Jia, C. Wen, D. Li, and M. Li, “Multi-topology contrastive graph representation learning,” Sci. China Inf. Sci., vol. 69, no. 2, pp. 122102:1–122102:11, 2026.
[23] Z. Zhang, H. Zhang, B. Wang, and J. Zhu, “Social graph diffusion and disentangled representations learning for multimodal recommendation,” J. King Saud Univ. Comput. Inf. Sci., 2026.
[24] S. Rendle, C. Freudenthaler, Z. Gantner, and L. Schmidt-Thieme, “BPR: Bayesian personalized ranking from implicit feedback,” in Proc. 25th Conf. Uncertainty in Artif. Intell. (UAI), 2009, pp. 452–461.
[25] L. Wu, J. Li, P. Sun, R. Hong, Y. Ge, and M. Wang, “Diffnet++: A neural influence and interest diffusion network for social recommendation,” IEEE Trans. Knowl. Data Eng., vol. 34, no. 10, pp. 4753–4766, 2020.
[26] J. Yu, H. Yin, X. Xia, T. Chen, L. Cui, and Q. V. Nguyen, “Are graph augmentations necessary? Simple graph contrastive learning for recommendation,” in Proc. 45th Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, 2022, pp. 1294–1303.
[27] L. Wu et al., “DcRec: Dual contrastive learning for social recommendation,” IEEE Trans. Knowl. Data Eng., 2022.
[28] T. Wang, L. Xia, and C. Huang, “Denoised self-augmented learning for social recommendation,” in Proc. 32nd Int. Joint Conf. Artif. Intell., Palo Alto, CA, USA: AAAI Press, 2022, pp. 2324–2331.
[29] J. Ding, G. Yu, X. He, Y. Xiang, and Y. Shen, “DENS: Disentangled negative sampling for contrastive learning in recommendation,” IEEE Trans. Knowl. Data Eng., 2023.
[30] B. Wu, L. Zhong, L. Yao, and Y. Ye, “EAGCN: An efficient adaptive graph convolutional network for item recommendation in social internet of things,” IEEE Internet Things J., vol. 9, no. 17, pp. 16386–16401, 2022.
پژوهش های نظری و کاربردی هوش ماشینی

مقالات آماده انتشار، پذیرفته شده
انتشار آنلاین از تاریخ 20 تیر 1405

فایل ها

هم رسانی

ارجاع به این مقاله

آمار