Class-Grouped Normalized Momentum and Faster Hyperparameter Exploration to Tackle Class Imbalance in Federated Learning
arXiv:2607.01474v1 Announce Type: new Abstract: Class imbalance poses a critical challenge in federated learning (FL), where underrepresented classes suffer from poor predictive performance yet cannot be addressed by standard centralized techniques due to privacy and heterogeneity constraints. We propose FedCGNM (Federated Class-Grouped Normalized Momentum), a client-side optimizer in FL that partitions classes into a small number of groups based on minimum within-group variance, maintains a momentum per group, normalizes each group momentum to unit length, and uses the summation of the normal
The proliferation of federated learning in privacy-sensitive applications necessitates robust solutions for common machine learning challenges like class imbalance, which standard centralized techniques cannot address.
Improving the performance of federated learning on real-world, imbalanced datasets expands its applicability across various industries, including healthcare and finance, where data privacy is paramount.
This new optimization method, FedCGNM, offers a practical approach to enhance the predictive accuracy of federated models, particularly for underrepresented classes, without compromising data privacy.
- · Federated learning developers
- · Healthcare sector
- · Financial services
- · Privacy-focused AI applications
- · Centralized machine learning approaches
- · Organizations with highly imbalanced federated datasets
Wider and more effective deployment of federated learning in production environments.
Increased trust and adoption of AI systems that operate on distributed and sensitive data.
Acceleration of privacy-preserving AI as a default standard across critical sectors.
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