π€ AI Summary
This work addresses the vulnerability of federated learning to redundant or anomalous client samples under non-IID and high-noise conditions, which often degrades model performance. The authors propose a federated sample selection method based on a multi-task autoencoder, coordinated by the central server to perform unsupervised anomaly detection using One-Class SVM, Isolation Forest, and an adaptive loss threshold. Notably, they introduce for the first time a federated multi-class Deep SVDD loss to enhance sample filtering in the feature space. Experimental results demonstrate that the proposed approach improves accuracy by up to 7.02% on CIFAR-10 and 1.83% on MNIST, with the federated SVDD component contributing an additional 0.99% gain, significantly boosting model robustness across varying client scales and noise levels.
π Abstract
Federated learning is a machine learning paradigm in which multiple devices collaboratively train a model under the supervision of a central server while ensuring data privacy. However, its performance is often hindered by redundant, malicious, or abnormal samples, leading to model degradation and inefficiency. To overcome these issues, we propose novel sample selection methods for image classification, employing a multitask autoencoder to estimate sample contributions through loss and feature analysis. Our approach incorporates unsupervised outlier detection, using one-class support vector machine (OCSVM), isolation forest (IF), and adaptive loss threshold (AT) methods managed by a central server to filter noisy samples on clients. We also propose a multi-class deep support vector data description (SVDD) loss controlled by a central server to enhance feature-based sample selection. We validate our methods on CIFAR10 and MNIST datasets across varying numbers of clients, non-IID distributions, and noise levels up to 40%. The results show significant accuracy improvements with loss-based sample selection, achieving gains of up to 7.02% on CIFAR10 with OCSVM and 1.83% on MNIST with AT. Additionally, our federated SVDD loss further improves feature-based sample selection, yielding accuracy gains of up to 0.99% on CIFAR10 with OCSVM. These results show the effectiveness of our methods in improving model accuracy across various client counts and noise conditions.