论文分享 | 推荐系统相关研究进展

我们从2024-12-02到2024-12-12的32篇文章中精选出5篇优秀的工作分享给读者。

1. PRECISE: Pre-training Sequential Recommenders with Collaborative and Semantic Information
2. Oracle-guided Dynamic User Preference Modeling for Sequential Recommendation
3. Temporal Linear Item-Item Model for Sequential Recommendation
4. BGTplanner: Maximizing Training Accuracy for Differentially Private Federated Recommenders via Strategic Privacy Budget Allocation
5. User-item fairness tradeoffs in recommendations

1.PRECISE: Pre-training Sequential Recommenders with Collaborative and Semantic Information

Authors: Chonggang Song, Chunxu Shen, Hao Gu, Yaoming Wu, Lingling Yi, Jie Wen, Chuan Chen

https://arxiv.org/abs/2412.06308

论文摘要

Real-world recommendation systems commonly offer diverse content scenarios for users to interact with. Considering the enormous number of users on industrial platforms, it is infeasible to utilize a single unified recommendation model to meet the needs of all scenarios. Typically, separate recommendation pipelines are established for each distinct scenario, which leads to challenges in comprehensively understanding users' interests. Recent research endeavors have been made to address this issue by pre-training models to encapsulate the overall interests of users. Traditional pre-trained recommendation models capture user interests by leveraging collaborative signals. However, a prevalent drawback of these systems is their incapacity to handle long-tail items and cold-start scenarios. With the recent advent of large language models (LLMs), there has been a significant increase in research efforts focused on using LLMs to extract semantic information for users and items. Nevertheless, text-based recommendations heavily rely on elaborate feature engineering and often fail to capture collaborative similarities.

To overcome these limitations, we propose a novel pre-training framework for sequential recommendation, termed PRECISE.. This framework combines collaborative signals with semantic information. Moreover, PRECISE employs a learning framework that initially models users' comprehensive interests across all recommendation scenarios and subsequently concentrates on the specific interests of target-scene behaviors. We demonstrate that PRECISE.~precisely captures the entire range of user interests and effectively transfers them to the target interests. Additionally, we introduce practical training strategies that enhance the model's performance in real-world applications. Empirical findings reveal that the PRECISE.~framework achieves outstanding performance on both public and industrial datasets. PRECISE.~has been deployed in multiple online recommendation scenarios within WeChat, and online A/B tests demonstrate substantial improvements in core business metrics.

论文简评

这篇论文深入探讨了如何通过集成语言模型中的协作信号和语义信息来改进传统推荐系统中用户兴趣表示和推荐性能的问题。该研究提出了一个名为PRECISE的框架，旨在解决冷启动和长尾场景中的推荐问题。论文详细介绍了PRECISE框架的三个核心模块——嵌入融合、通用训练和目标化训练，这些模块通过有效的方式整合了协同信号和语义信息，从而显著提高了推荐系统的性能。此外，论文提供了大量实证研究结果，证明了PRECISE方法的有效性与优越性，特别是在公共数据集和工业数据集上的表现。总的来说，这篇论文不仅为推荐系统的研究提供了一种创新的方法论，也为实际应用提供了有效的解决方案。

2.Oracle-guided Dynamic User Preference Modeling for Sequential Recommendation

Authors: Jiafeng Xia, Dongsheng Li, Hansu Gu, Tun Lu, Peng Zhang, Li Shang, Ning Gu

https://arxiv.org/abs/2412.00813

论文摘要

Sequential recommendation methods can capture dynamic user preferences from user historical interactions to achieve better per formance. However, most existing methods only use past informa tion extracted from user historical interactions to train the models, leading to the deviations of user preference modeling. Besides past information, future information is also available during training, which contains the “oracle” user preferences in the future and will be beneficial to model dynamic user preferences. Therefore, we pro pose an oracle-guided dynamic user preference modeling method for sequential recommendation (Oracle4Rec), which leverages fu ture information to guide model training on past information, aim ing to learn “forward-looking” models. Specifically, Oracle4Rec f irst extracts past and future information through two separate encoders, then learns a forward-looking model through an oracle guiding module which minimizes the discrepancy between past and future information. We also tailor a two-phase model training strategy to make the guiding more effective. Extensive experiments demonstrate that Oracle4Rec is superior to state-of-the-art sequen tial methods. Further experiments show that Oracle4Rec can be leveraged as a generic module in other sequential recommendation methods to improve their performance with a considerable margin.

论文简评

这篇论文是关于推荐系统研究的重要成果。该文提出了一种名为Oracle4Rec的方法，它利用历史交互数据与未来交互数据相结合来提高模型性能。本文的主要优点在于提出了一个分离的过去信息和未来信息的编码器，并引入了一个用于最小化两者之间差异的Oracle引导模块。

论文的核心是通过实验展示了其对现有方法的有效性。作者在多个数据集上进行了详细的实验，结果表明Oracle4Rec在提升模型准确性方面取得了显著成效。此外，该文还强调了Oracle4Rec的通用性和可扩展性，表明它不仅可以应用于其他序列算法中，而且可以作为一个通用模块被广泛应用。

总的来说，《Oracle4Rec》是一篇非常有价值的学术论文，它不仅为推荐系统的研究提供了新的视角，也展示了其在实际应用中的强大能力。

3.Temporal Linear Item-Item Model for Sequential Recommendation

Authors: Seongmin Park, Mincheol Yoon, Minjin Choi, Jongwuk Lee

https://arxiv.org/abs/2412.07382

论文摘要

In sequential recommendation (SR), neural models have been actively explored due to their remarkable performance, but they suffer from inefficiency inherent to their complexity. On the other hand, linear SR models exhibit high efficiency and achieve competitive or superior accuracy compared to neural models. However, they solely deal with the sequential order of items (i.e., sequential information) and overlook the actual timestamp (i.e., temporal information). This limitation affects their ability to effectively capture various user preference drifts over time. To address this issue, we propose a novel linear SR model named the TemporAl LinEar item-item model (TALE), which incorporates temporal information while preserving training and inference efficiency through three key components: (i) Single-target augmentation concentrates on a single target item, enabling us to learn the temporal correlation for the target item. (ii) Time interval-aware weighting utilizes the actual timestamp to discern item correlation depending on time intervals. (iii) Trend-aware normalization reflects the dynamic shift of item popularity over time. Our empirical studies show that TALE outperforms ten competing SR models with gains of up to 18.71% on five benchmark datasets and exhibits remarkable effectiveness in evaluating long-tail items with improvements of up to 30.45%. The source code is available at https://github.com/psm1206/TALE.

论文简评

这篇论文介绍了Temporal Linear Item-Item（TALE）模型，旨在在时间序列推荐系统中结合时序信息与计算效率。TALE模型集成了时间信息，并保持了计算效率，包含针对单目标增强、基于时间间隔加权、以及趋势感知正则化的组件。实验证明，TALE模型在多个数据集上在准确性和效率方面均优于竞争对手，特别是在处理长尾项时表现更为突出。此外，论文还强调了训练和推理过程中的高效性，相比于神经网络模型，TALE在时间和计算成本上实现了显著节省。总的来说，这篇论文为时间序列推荐系统的改进具有重要意义，提供了处理时序数据的新的思路和解决方案。

4.BGTplanner: Maximizing Training Accuracy for Differentially Private Federated Recommenders via Strategic Privacy Budget Allocation

Authors: Xianzhi Zhang, Yipeng Zhou, Miao Hu, Di Wu, Pengshan Liao, Mohsen Guizani, Michael Sheng

https://arxiv.org/abs/2412.02934

论文摘要

To mitigate the rising concern about privacy leakage, the federated recommender (FR) paradigm emerges, in which decentralized clients co-train the recommendation model without exposing their raw user-item rating data. The differentially private federated recommender (DPFR) further enhances FR by injecting differentially private (DP) noises into clients. Yet current DPFRs suffer from noise distortion, which prevents them from achieving satisfactory accuracy. Various efforts have been made to improve DPFRs by adaptively allocating the privacy budget throughout the learning process. However, due to the intricate relationship between privacy budget allocation and model accuracy, existing works are still far from maximizing DPFR accuracy. To address this challenge, we develop BGTplanner (Budget Planner) to strategically allocate the privacy budget for each round of DPFR training, thereby improving overall training performance. Specifically, we leverage Gaussian process regression and historical information to predict how changes in the allocated privacy budget affect recommendation accuracy. Additionally, Contextual Multi-Armed Bandit (CMAB) is employed to make privacy budget allocation decisions by reconciling current improvements with long-term privacy constraints. Our extensive experimental results on real datasets demonstrate that BGTplanner achieves an average improvement of 6.76% in training performance compared to state-of-the-art baselines.

论文简评

BGTplanner 是一个在线算法，旨在通过策略性分配隐私预算来最大化不同隐私保护联邦推荐系统的训练精度。该文结合高斯过程回归预测学习改进，并利用上下文多臂老虎机进行决策制定。该方法旨在解决隐私保护与模型准确性之间的平衡问题，特别是在分布式学习环境中。

文章的关键点在于解决了分布式学习中的重要挑战，引入了一种新颖的方法，并结合高斯过程回归和上下文多臂老虎机，为隐私保护的DPFR提供了新的视角，展示了与现有方法相比的显著性能提升。

5.User-item fairness tradeoffs in recommendations

Authors: Sophie Greenwood, Sudalakshmee Chiniah, Nikhil Garg

https://arxiv.org/abs/2412.04466

论文摘要

In the basic recommendation paradigm, the most (predicted) relevant item is recommended to each user. This may result in some items receiving lower exposure than they "should"; to counter this, several algorithmic approaches have been developed to ensure item fairness. These approaches often lead to degraded recommendations for some users in order to improve item outcomes, raising user fairness concerns. Recently, research has focused on developing algorithms for multi-sided fairness to jointly optimize user fairness, item fairness, and overall recommendation quality. This induces the question: what is the tradeoff between these objectives, and what are the characteristics of optimal solutions? Theoretically, we develop a model of recommendations with user and item fairness objectives and characterize the solutions of fairness-constrained optimization. We identify two phenomena: (a) when user preferences are diverse, there is "free" item and user fairness; and (b) users whose preferences are misestimated can be especially disadvantaged by item fairness constraints. Empirically, we prototype a recommendation system for preprints on arXiv and implement our framework, measuring these phenomena in practice and showing how they inform the design of markets with recommendation systems-intermediated matching.

论文简评

这篇关于推荐系统公平性问题的研究论文，主要探讨了用户公平性和物品公平性的平衡，并提出了一个理论框架来解决多边公平目标。研究者通过实证分析，使用一个实际原型推荐系统评估了对阿克塞尔文预印本的影响，以验证其理论发现。论文整体结构清晰，内容丰富且富有创新性，尤其针对当前推荐系统中普遍存在的公平性问题提供了新的视角和解决方案。此外，论文还强调了多方公平性的重要性，以及如何通过改进算法和技术实现这一目标。总之，该论文为提高推荐系统的公正性和透明度提供了一种可行的方法，并为相关领域的研究提供了有价值的参考。

我们欢迎您在评论区中留下宝贵的建议！包括但不限于：

• 可以提出推文中论文简评的不足！
• 可以分享最近更值得推荐的论文并给出理由！

推荐阅读