In recent years, two major issues in recommendation systems have received extensive attention from researchers: (1) The incompleteness of user interaction data: In recommendation systems, user interaction data is often incomplete. (2) Difficulty in accurately reflecting users' true preferences through interactions: User interaction data may suffer from selection biases, with some interactions possibly being noise, leading to misinterpretation of user preferences by the recommendation system. To address these challenges, we propose an Attention-based Multilayer Linear Diffusion Model (AMLDM). Specifically, we gradually introduce pre-defined Gaussian noise into the forward process to disrupt users' interaction histories. Subsequently, through multiple attention-based linear layers, we iteratively restore the damaged interaction histories incurred during th e forward process, ensuring that the distribution of the repaired interaction history aligns with the original distribution of user interaction history. By injecting an appropriate amount of noise into users' interaction histories, we enhance the robustness of the recommendation system and learn users' true preferences during the reverse process. Comparative analysis with several benchmark recommendation system models demonstrates the significant advantages of our proposed algorithm in recommendation performance.

Yixuan Zhang. Attention-based Multilayer Linear Diffusion Model. Academic Journal of Computing & Information Science (2024), Vol. 7, Issue 4: 56-61. https://doi.org/10.25236/AJCIS.2024.070408.

[1] Liang D, Krishnan R G, Hoffman M D, et al. Variational autoencoders for collaborative filtering[C]// Proceedings of the 2018 world wide web conference. 2018: 689-698.

[2] Shenbin I, Alekseev A, Tutubalina E, et al. Recvae: A new variational autoencoder for top-n recommendations with implicit feedback[C]//Proceedings of the 13th international conference on web search and data mining. 2020: 528-536.

[3] 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. 2020: 639-648.

[4] 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. 2020: 1001-1010.

[5] Rendle S, Freudenthaler C, Gantner Z, et al. BPR: Bayesian personalized ranking from implicit feedback [J]. arXiv preprint arXiv:1205.2618, 2012.