analiza aiqbench
This arXiv paper proposes using interaction timestamps to identify and downweight noisy edges in neural graph collaborative filtering, a common approach to recommendation. The two methods, Tint-A (reweighting the adjacency matrix) and Tint-L (reweighting the loss), are evaluated on four real-world datasets. For ML practitioners, the main takeaway is that temporal signals can be a practical, low-cost complement to embedding-based denoising, though the results are benchmark-specific and not yet validated in production systems.
Time Will Tell: Harness Time to Denoise Collaborative Filtering
arXiv:2411.09181v2 Announce Type: replace-cross Abstract: Neural graph collaborative filtering is a powerful tool for recommendation due to its superb capability of capturing the user-item interaction. Despite the substantial progress, the noisy nature of interaction data remains a significant challenge in collaborative filtering, especially for data collected from implicit feedback. Both noisy interactions and interactions crafted by malicious users could greatly degenerate model performance. To purify the input data and enhance model robustness, it is vital to understand how noisy interactions can be identified and rectified. Existing denoising methods focus on mining the correlation between user/item embeddings and noisy interactions, neglecting the temporal dimension for denoising. However, time information bears its natural and unique advantage in revealing noisy edges as we observe in this paper. In this work, we introduce a simple yet effective mechanism for generating time-aware user/item embeddings that capture both global preferences and temporal patterns. Two strategies, reweighting adjacency matrix (Tint-A) and reweighting loss function (Tint-L), are proposed to leverage the embeddings for the task of using \underline{t}ime \underline{in}forma\underline{t}ion to denoise \underline{c}ollaborative \underline{f}iltering (TintCF). Extensive experiments have been conducted on four real-world datasets to demonstrate the efficacy of our methods and illustrate how time information indeed helps identify and rectify noisy edges. Our codes are available at \href{https://github.com/xinyuu-he/TintCF}{https://github.com/xinyuu-he/TintCF}.
Abstract:Neural graph collaborative filtering is a powerful tool for recommendation due to its superb capability of capturing the user-item interaction. Despite the substantial progress, the noisy nature of interaction data remains a significant challenge in collaborative filtering, especially for data collected from implicit feedback. Both noisy interactions and interactions crafted by malicious users could greatly degenerate model performance. To purify the input data and enhance model robustness, it is vital to understand how noisy interactions can be identified and rectified. Existing denoising methods focus on mining the correlation between user/item embeddings and noisy interactions, neglecting the temporal dimension for denoising. However, time information bears its natural and unique advantage in revealing noisy edges as we observe in this paper. In this work, we introduce a simple yet effective mechanism for generating time-aware user/item embeddings that capture both global preferences and temporal patterns. Two strategies, reweighting adjacency matrix (Tint-A) and reweighting loss function (Tint-L), are proposed to leverage the embeddings for the task of using \underline{t}ime \underline{in}forma\underline{t}ion to denoise \underline{c}ollaborative \underline{f}iltering (TintCF). Extensive experiments have been conducted on four real-world datasets to demonstrate the efficacy of our methods and illustrate how time information indeed helps identify and rectify noisy edges. Our codes are available at \href{this https URL}{this https URL}.
Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2411.09181 [cs.IR]
(or arXiv:2411.09181v2 [cs.IR] for this version)
https://doi.org/10.48550/arXiv.2411.09181 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 307 cuvinte
Subiecte conexe
X-KGRank: A Knowledge Graph RAG Framework for Explainable Recommendations via Pattern Mining and LLM Re-Ranking
arXiv · 5 aug. 2026
GSPRec: On Improving Item Representations in Graph Signal Processing for Collaborative Filtering
arXiv · 23 iul. 2026
A Multi-Source Ensemble Approach to Candidate Generation for Alternative Vacation Rental Property Recommendations
arXiv · 9 sept. 2026
Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation
arXiv · 6 iun. 2026