Machine learning faces challenges from incomplete and biased data
Machine Learning under Imperfect Data: Challenges and Methods
Machine Learning
Summary
Machine learning models often assume that the data they learn from is complete, balanced, and clean, but real-world data usually isn't. This paper reviews common problems like missing measurements, rare categories, and changes between training and deployment data. The authors group these issues into shared causes such as loss of information and unclear supervision, and they look at methods to handle them like data reconstruction, balancing, and adapting to new environments. They also discuss the limits of fixing imperfect data and suggest future directions for more reliable learning and evaluation.
What this means in practice
- •For data engineering teams: Improve data pipelines by applying methods to reconstruct and rebalance datasets with missing or biased data.
- •For software engineers: Build machine learning models that better handle changes in deployment environments through domain adaptation techniques.
A survey. It maps existing work.
Authors
Masoumeh Zareapoor
Abstract
Machine-learning models are commonly developed under an assumption that training and test data are sufficiently complete, balanced, labelled, and drawn from compatible distributions. In practice, one or more of these conditions is often violated. Measurements may be missing or corrupted, rare classes may be poorly represented, supervision may be weak, and the deployment environment may differ from the training environment. These imperfections are usually treated as separate technical problems, although they alter learning through a small number of shared mechanisms: loss of information, biased empirical risk, ambiguous supervision, and unstable representations. This short survey organises representative methods around these mechanisms. It reviews reconstruction and generation, rebalancing and representation calibration, learning with limited supervision, adaptation across domains and modalities, and reliability under distribution change. The discussion highlights the limits of plausible reconstruction, benchmark-specific correction, and adaptation without trustworthy feedback. It concludes with directions for evidence-aware learning, uncertainty-preserving prediction, and evaluation that separates visual plausibility from decision utility.