WHERE to Generate Matters: Budget-Aware Synthetic Augmentation for Label Skewed Federated Learning
arXiv:2607.06616v1 Announce Type: new Abstract: Label skew in federated learning (FL) causes client drift and degrades global accuracy. Synthetic data augmentation can reduce this imbalance; however, full class balancing requires substantial computation cost. We propose FedEAS, a policy that assigns each client an entropy-adaptive per-class generation budget computed from its local label distribution. The budget jointly decides \emph{how much} each client generates and \emph{WHERE} the samples go. Accordingly, the total generation budget follows from the per-client budgets rather than being fi
The increasing scale and complexity of AI models, coupled with growing privacy concerns, demand more efficient and less data-intensive training methods like federated learning.
This research addresses a critical limitation in federated learning, making it more robust and applicable to real-world scenarios where data distribution is inherently imbalanced.
The ability to perform budget-aware synthetic augmentation in federated learning will lead to more accurate AI models on diverse datasets without requiring centralized data collection.
- · AI developers
- · Healthcare sector
- · Privacy-focused industries
- · Edge computing providers
- · Traditional centralized data training approaches
Improved performance and broader adoption of federated learning in various applications are direct outcomes.
This could enable new AI services where data privacy is paramount, such as in medical diagnostics or personalized recommendations.
Reduced computational costs and energy footprint for training large AI models may indirectly contribute to more sustainable AI development.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG