Unified Analysis of Continuous Weak Features Learning with Applications to Learning from Missing Data
Kosuke Sugiyama, Masato Uchida
Abstract
This paper addresses weak features learning (WFL), focusing on learning scenarios characterized by low-quality input features (weak features; WFs) that arise due to missingness, measurement errors, or ambiguous observations. We present a theoretical formalization and error analysis of WFL for continuous WFs (continuous WFL), which has been insufficiently explored in existing literature. A previous study established formalization and error analysis for WFL with discrete WFs (discrete WFL); however, this analysis does not extend to continuous WFs due to the inherent constraints of discreteness. To address this, we propose a theoretical framework specifically designed for continuous WFL, systematically capturing the interactions between feature estimation models for WFs and label prediction models for downstream tasks. Furthermore, we derive the theoretical conditions necessary for both sequential and iterative learning methods to achieve consistency. By integrating the findings of this study on continuous WFL with the existing theory of discrete WFL, we demonstrate that the WFL framework is universally applicable, providing a robust theoretical foundation for learning with low-quality features across diverse application domains.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
1 top-tier paper has cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- What's a good imputation to predict with missing values? cites NeuMiss networks: differentiable programming for supervised learning with missing values
- A Unified Framework for Generalization Error Analysis of Learning with Arbitrary Discrete Weak Features cites What's a good imputation to predict with missing values?
- A Unified Framework for Generalization Error Analysis of Learning with Arbitrary Discrete Weak Features cites NeuMiss networks: differentiable programming for supervised learning with missing values
- A Unified Framework for Generalization Error Analysis of Learning with Arbitrary Discrete Weak Features cites Conformal Prediction with Missing Values
- Conformal Prediction with Missing Values cites What's a good imputation to predict with missing values?
- Conformal Prediction with Missing Values cites NeuMiss networks: differentiable programming for supervised learning with missing values
Cited by top-tier papers1
Builds on4
- What's a good imputation to predict with missing values?Marine Le Morvan, Julie Josse, Erwan Scornet, Gaël VaroquauxNeurIPS 2021 · 95 citations
- NeuMiss networks: differentiable programming for supervised learning with missing valuesMarine Le Morvan, Julie Josse, Thomas Moreau, Erwan Scornet et al.NeurIPS 2020 · 50 citations
- Conformal Prediction with Missing ValuesMargaux Zaffran, Aymeric Dieuleveut, Julie Josse, Yaniv RomanoICML 2023 · 31 citations
- A Unified Framework for Generalization Error Analysis of Learning with Arbitrary Discrete Weak FeaturesKosuke Sugiyama, Masato UchidaICML 2025
Related papers
- Partial-Label RegressionXin Cheng, Deng-Bao Wang, Lei Feng, Min-Ling Zhang et al.AAAI 2023 · 4 citations
- ULAREF: A Unified Label Refinement Framework for Learning with Inaccurate SupervisionCongyu Qiao, Ning Xu, Yihao Hu, Xin GengICML 2024 · 1 citation
- Universalizing Weak SupervisionChangho Shin, Winfred Li, Harit Vishwakarma, Nicholas Carl Roberts et al.ICLR 2022 · 35 citations
- Generative Modeling Reinvents Supervised Learning: Label Repurposing with Predictive Consistency LearningYang Li, Jiale Ma, Yebin Yang, Qitian Wu et al.ICML 2025
- Structured Prediction with Partial Labelling through the Infimum LossVivien Cabannes, Alessandro Rudi, Francis R. BachICML 2020 · 50 citations
