Federated learning signals reveal different types of data corruption
Different Corruptions, Different Signals: Uncertainty and Loss in Federated Data Quality
Machine LearningArtificial IntelligenceComputer Vision and Pattern Recognition
Summary
Data in federated learning can be corrupted either by noisy inputs (like blurry images) or wrong labels (misnamed pictures). This paper shows that two ways to detect these problems—measuring uncertainty in predictions and checking prediction errors against labels—work differently depending on the corruption type. Prediction errors help spot wrong labels, while uncertainty measures better detect noisy inputs. The authors suggest choosing the right detection method based on the kind of corruption to improve data quality in federated learning.
What this means in practice
- •For machine learning engineers: Improve federated learning model quality by using prediction loss signals to detect label errors and uncertainty measures to detect noisy inputs across clients.
- •For distributed system operators: Monitor data corruption types in federated systems to prioritize fixing label errors or input noise according to the different detection signals described.
Authors
Bradley Scott, Zeqi Luo, Edmond S. L. Ho
Abstract
Federated learning (FL) data corruption can affect either inputs or labels, but it remains unclear whether input-conditional uncertainty and prediction-label loss expose these corruption modes equally. This paper compares two corruption-detection signals in FL: input-conditional uncertainty and prediction-label loss. The uncertainty signal is characterised using a learned aleatoric variance estimate together with Monte Carlo (MC) dropout variance and entropy measures, while the loss is computed against the supplied label. We test these signals against additive image noise and persistent random label flips. On ResNet-20 with CIFAR-10 and SVHN under Dirichlet partitions with data that are not independent and identically distributed (non-IID), the two corruption types behave differently. For persistent random label flips, the within-client per-sample area under the receiver operating characteristic curve (AUC) is 0.85 on CIFAR-10 and 0.95 on SVHN for prediction-label loss, while every uncertainty estimator stays at chance (0.49--0.50). This pattern is consistent with the model remaining confident in the underlying image despite the supplied label being wrong. For image noise, expected-entropy uncertainty rises above chance (0.67 on CIFAR-10 and 0.66 on SVHN), while loss responds comparably (0.64 on both). Each signal is therefore the stronger detector for a different corruption: the prediction-label loss for persistent label flips, and expected-entropy uncertainty for image noise, with its advantage becoming apparent as federation-wide corruption prevalence increases. Robust FL data-quality assessment should match the signal to the corruption rather than rely on uncertainty alone across corruption types.