AI summaryⓘ
The authors focus on Personalized Federated Learning (PFL), where the goal is to create custom models for different groups instead of one model for everyone. They point out that previous methods rely on repeated model updates to group users, which can be easily messed up by bad data or wrong labels. To fix this, they propose FB-NLL, which groups users by analyzing the structure of their data features before training starts, avoiding problems caused by noisy updates and reducing computation. They also offer a way to spot and fix incorrect labels within groups by checking feature consistency, without complicated noise estimation. Their method works with various models and improves accuracy and stability across different datasets and noisy conditions.
Personalized Federated Learninguser clusteringheterogeneous datafeature spacespectral structuresubspace similaritynoisy labelslabel correctionnoise-robust trainingmodel-independent
Authors
Abdulmoneam Ali, Ahmed Arafa
Abstract
Personalized Federated Learning (PFL) aims to learn multiple task-specific models rather than a single global model across heterogeneous data distributions. Existing PFL approaches typically rely on iterative optimization-such as model update trajectories-to cluster users that need to accomplish the same tasks together. However, these learning-dynamics-based methods are inherently vulnerable to low-quality data and noisy labels, as corrupted updates distort clustering decisions and degrade personalization performance. To tackle this, we propose FB-NLL, a feature-centric framework that decouples user clustering from iterative training dynamics. By exploiting the intrinsic heterogeneity of local feature spaces, FB-NLL characterizes each user through the spectral structure of the covariances of their feature representations and leverages subspace similarity to identify task-consistent user groupings. This geometry-aware clustering is label-agnostic and is performed in a one-shot manner prior to training, significantly reducing communication overhead and computational costs compared to iterative baselines. Complementing this, we introduce a feature-consistency-based detection and correction strategy to address noisy labels within clusters. By leveraging directional alignment in the learned feature space and assigning labels based on class-specific feature subspaces, our method mitigates corrupted supervision without requiring estimation of stochastic noise transition matrices. In addition, FB-NLL is model-independent and integrates seamlessly with existing noise-robust training techniques. Extensive experiments across diverse datasets and noise regimes demonstrate that our framework consistently outperforms state-of-the-art baselines in terms of average accuracy and performance stability.