Double Correction Framework for Denoising Recommendation
School of Computer Science and Information Engineering, Hefei University of Technology(合肥工业大学计算机与信息工程学院);Department of Computer Science and Technology, Tsinghua University(清华大学计算机科学与技术系);Institute of Dataspace, Hefei Comprehensive National Science Center(合肥综合性国家科学中心空间研究院);Department of Mathematics, University of Macau(澳门大学数学系)
Zhuangzhuang He, Yifan Wang, Yonghui Yang, Peijie Sun, Le Wu, Haoyue Bai, Jinqi Gong, Richang Hong, Min Zhang
As its availability and generality in online services, implicit feedback is more commonly used in recommender systems. However, implicit feedback usually presents noisy samples in real-world recommendation scenarios (such as misclicks or non-preferential behaviors), which will affect precise user preference learning. To overcome the noisy samples problem, a popular solution is based on dropping noisy samples in the model training phase, which follows the observation that noisy samples have higher training losses than clean samples. Despite the effectiveness, we argue that this solution still has limits. (1) High training losses can result from model optimization instability or hard samples, not just noisy samples. (2) Completely dropping of noisy samples will aggravate the data sparsity, which lacks full data exploitation. To tackle the above limitations, we propose a Double Correction Framework for Denoising Recommendation (DCF), which contains two correction components from views of more precise sample dropping and avoiding more sparse data. In the sample dropping correction component, we use the loss value of the samples over time to determine whether it is noise or not, increasing dropping stability. Instead of averaging directly, we use the damping function to reduce the bias effect of outliers. Furthermore, due to the higher variance exhibited by hard samples, we derive a lower bound for the loss through concentration inequality to identify and reuse hard samples. In progressive label correction, we iteratively re-label highly deterministic noisy samples and retrain them to further improve performance. Finally, extensive experimental results on three datasets and four backbones demonstrate the effectiveness and generalization of our proposed framework.
Denoising Implicit Feedback for Cold-start Recommendation
Hong Kong Baptist University(香港浸会大学);Independent Researcher(独立研究员);Peking University(北京大学);Nanjing University(南京大学);Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所)
Burger: Robust Graph Denoising-augmentation Fusion and Multi-semantic Modeling in Social Recommendation
Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing(未来区块链与隐私计算先进创新中心);School of Artificial Intelligence, Beihang University(北京航空航天大学人工智能学院);School of Cyber Science and Engineering, Huazhong University of Science and Technology(华中科技大学网络安全与工程学院)
Enhancing Recommendation with Denoising Auxiliary Task
College of Big Data and information engineering, Guizhou University(贵州大学大数据与信息工程学院);Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所);University of the Chinese Academy of Sciences(中国科学院大学)
Learning Robust Recommenders through Cross-Model Agreement
University of Science and Technology of China(中国科学技术大学);Shandong University(山东大学);Cambridge University(剑桥大学);University of Glasgow(格拉斯哥大学);University of Science of Technology of China(中国科学技术大学)
Denoising Long- and Short-term Interests for Sequential Recommendation
Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所);College of Computer Science, Chongqing University(重庆大学计算机学院);University of Chinese Academy of Sciences(中国科学院大学)
A Missing Information Loss function for implicit feedback datasets
BBVA Data & Analytics(BBVA数据分析公司)
隐式反馈推荐系统的潜在因素模型通常将未观察到的用户-项目交互(即丢失信息)视为负反馈。这通常是通过负采样(逐点损失)或排序损失函数(成对或列表估计)来完成的。由于零偏好推荐是大多数常见目标函数的有效解决方案,因此将未知值视为实际零会导致用户对大多数可用项目具有零偏好推荐。在本文中,我们提出了一种新的目标函数,\emph{Missing Information Loss}(MIL),明确禁止治疗未观察到的用户项交互的积极或消极的反馈。我们将这种损失应用于传统的矩阵分解和基于用户的去噪自动编码器,并将其与其他已建立的目标函数进行比较,如交叉熵(点和成对)或最近提出的多项式对数似然。MIL在应用于三个数据集时,在排名感知指标方面取得了有竞争力的性能。此外,我们表明,这种相关性的建议,同时显示受欢迎的项目不太频繁(最多$20 \%$减少相对于最好的竞争方法)。这种对流行项目推荐的去偏倾向于不常见项目的出现(直到长尾推荐的$50 \%$增加),这对于具有大的产品目录的推荐系统是有价值的特征。
Dynamic Spectral Denoising with Global-Context Attention for Multi-Behavior Recommendation
National University of Singapore(新加坡国立大学);Singapore Management University(新加坡管理大学);Hefei University of Technology(合肥工业大学);University of Science and Technology of China(中国科学技术大学)