Missing data is a pervasive challenge in machine learning, often degrading predictive performance and complicating model deployment. Most existing approaches treat imputation as a preprocessing step independent of the downstream task, a separation that becomes problematic under train-test mismatches in missingness mechanisms or rates. We introduce a unified framework that treats missing entries as variables in a task-conditioned optimization problem. We instantiate this framework in two settings: an amortized approach for scenarios with missing data at both training and test time, and a latent optimization approach for settings where only training data is incomplete.
The evaluation is conducted on 98 tabular datasets, and robustness under train-test mismatches is assessed on 36 datasets. Results reveal that reconstruction error constitutes a poor proxy for predictive utility (Spearman rho approximately 0.16 to 0.32), and that task-aware imputation becomes critical in scenarios where classical methods fail to recover complete-data performance, achieving win rates of up to 80% in challenging train-test mismatch scenarios.
DiffPuter: Empowering Diffusion Models for Missing Data Imputation
Computer Science Department, University of Illinois at Chicago(伊利诺伊大学芝加哥分校计算机科学系);Eric and Wendy Schmidt Center, Broad Institute of MIT and Harvard(MIT和哈佛大学Broad研究所埃里克和文迪·施密特中心)
Differentiable and Scalable Generative Adversarial Models for Data Imputation
Center for Data Science, Zhejiang University(浙江大学数据科学中心);Information Hub, The Hong Kong University of Science and Technology(香港科技大学信息枢纽);College of Computer Science, Zhejiang University(浙江大学计算机科学与技术学院)
Causal View of Time Series Imputation: Some Identification Results on Missing Mechanism
School of Computer Science, Guangdong University of Technology(广东技术大学计算机科学学院);Peng Cheng Laboratory(鹏城实验室);Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学);College of Science, Shantou University(汕头大学理学院)
Learnable Prompt as Pseudo-Imputation: Rethinking the Necessity of Traditional EHR Data Imputation in Downstream Clinical Prediction
Peking University(北京大学);School of Computer Science(计算机学院);School of Software and Microelectronics(软件与微电子学院);Center on Frontiers of Computing Studies(计算前沿研究中心);National Engineering Research Center for Software Engineering(国家软件工程工程研究中心);Peking University Information Technology Institute (Tianjin Binhai)(北京大学信息技术学院(天津滨海))
本文提出可学习提示作为伪插补(PAI)的新训练协议,以可学习提示建模下游模型对缺失值的隐式偏好,替代传统 EHR 插补模型,在多个临床预测任务和数据集上显著提升现有模型性能并具有更强鲁棒性。
One Pipeline, Many Transformers: Pattern-Specific Imputation Specialists for Tabular Missing Data
Industrial Engineering \& Operations Research, Columbia University, New York, USA;Department of Statistics, Columbia University, New York, USA;Operations Research \& Information Engineering, Cornell Tech, New York, USA
Is Precise Recovery Necessary? A Task-Oriented Imputation Approach for Time Series Forecasting on Variable Subset
University of Macau(澳门大学);Northeast Normal University(东北师范大学);Computer Network Information Center, Chinese Academy of Sciences(中国科学院计算机网络信息中心);University of Oxford(牛津大学);Dalian Maritime University(大连海事大学)
模型可以暴露关于其训练数据的敏感信息。在属性推断攻击中,攻击者对一些训练记录具有部分知识,并且可以访问在这些记录上训练的模型,并且推断出这些记录的敏感特征的未知值。我们研究了细粒度的属性推断的变体,我们称之为\emph{sensitive value inference},对手的目标是以高置信度从候选集的未知属性具有特定的敏感值的一些记录。我们明确比较属性推断与数据填补,捕捉训练分布统计,在各种假设下的训练数据提供给对手。我们的主要结论是:(1)先前的属性推断方法并没有比对手在不访问训练模型的情况下能够推断出的更多地揭示来自模型的训练数据,但是具有与训练属性推断攻击所需的相同的底层分布知识;(2)黑盒属性推断攻击很少学习没有模型就不能学习的任何东西;但是(3)白盒攻击,我们在本文中介绍和评估,可以可靠地识别具有敏感值属性的一些记录,这些属性在没有访问模型的情况下无法预测。此外,我们表明,提出的防御措施,如差异化私人培训和从培训中删除易受攻击的记录,并不能减轻这种隐私风险。我们的实验代码可在\url{https://github.com/bargavj/EvaluatingDPML}上获得。
Handling Missing Data in Downstream Tasks With Distribution-Preserving Guarantees
Institute of Computer Science University of Rostock(计算机科学研究所罗斯托克大学);Institute of Computer Science / Soda University of Rostock / Inria Saclay(计算机科学研究所/ Soda罗斯托克大学/ Inria萨克雷);Department of Data Science IISER Thiruvananthapuram(数据科学系IISER特里凡纳塔普拉姆)
针对分类等下游任务中的缺失数据问题,提出可与分类器联合训练的 F3I 迭代 K 近邻插补方法,并给出分布保持理论保证与实验验证。
Nonparametric End-to-End Probabilistic Forecasting of Distributed Generation Outputs Considering Missing Data Imputation
Guangzhou Power Supply Bureau Guangdong Power Grid Co., Ltd.(广东电网有限责任公司广州供电局);Tsinghua University(清华大学);Tsinghua-Berkeley Shenzhen Institute (TBSI)(清华-伯克利深圳学院)
Semi-Supervised Learning with Multiple Imputations on Non-Random Missing Labels
Millburn High School(米尔本高中);Franklin High School(富兰克林高中);Beijing University of Posts and Telecommunication(北京邮电大学);Yew Wah International Education School of Guangzhou(广州耀华国际教育学校)
Analyzing and Improving Diffusion Models for Time-Series Data Imputation: A Proximal Recursion Perspective
State Key Lab of General AI, School of Intelligence Science and Technology, Peking University(人工智能国家重点实验室、智能科学与技术学院、北京大学);Zhejiang University(浙江大学);Beijing Academy of Artificial Intelligence (BAAI)(北京人工智能研究院);The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳));Center for Data Science, Peking University(北京大学数据科学中心);Institute for Artificial Intelligence, Peking University(北京大学人工智能研究院);Pazhou Laboratory (Huangpu), Guangzhou, China(黄埔实验室(华南))
Temporally Multi-Scale Sparse Self-Attention for Physical Activity Data Imputation
University of Massachusetts Amherst(马萨诸塞大学阿默斯特分校);Georgia Institute of Technology(佐治亚理工学院);University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);University of Memphis(孟菲斯大学)
Extended Missing Data Imputation via GANs for Ranking Applications
Cornell University(康奈尔大学);Amazon(亚马逊)
我们提出了条件填补GAN,一种基于生成对抗网络(GANs)的扩展缺失数据填补方法。激励用例是学习排名,现代搜索,推荐系统和信息检索应用程序的基石。经验排序数据集并不总是遵循标准高斯分布或完全随机缺失(MCAR)机制,这是经典缺失数据插补方法的标准假设。我们的方法提供了一个简单的解决方案,提供兼容的插补保证,同时放松假设缺失的机制和回避近似棘手的分布,以提高插补质量。我们证明了最佳的GAN填补实现扩展随机缺失(EMAR)和扩展总是随机缺失(EAMAR)机制,超越了天真的MCAR。我们的方法在开源Microsoft Research Ranking(MSR)数据集和合成排名数据集上展示了最高的插补质量,与最先进的基准和各种特征分布相比。使用专有的Amazon Search排名数据集,我们还展示了与地面实况数据相比,在GAN估算数据上训练的排名模型的可比排名质量指标。