Advances in Federated Learning and Differential Privacy

The field of federated learning and differential privacy is rapidly evolving, with a focus on developing innovative methods to protect individual privacy while enabling collaborative model training. Recent research has explored the use of adaptive federated learning algorithms, client-level differential privacy, and distributionally robust optimization to improve model performance and robustness in heterogeneous environments. Notably, the development of personalized federated learning methods, such as those using particle-based variational inference and Wasserstein barycenter aggregation, has shown promising results in handling non-i.i.d. client data and quantifying uncertainty. Noteworthy papers include Optimal Allocation of Privacy Budget on Hierarchical Data Release, which formulates the challenge of optimal privacy budget allocation as a constrained optimization problem, and Heterogeneity-Aware Client Sampling, which proposes a unified solution for consistent federated learning by eliminating objective inconsistency. Additionally, the paper FedDuA: Doubly Adaptive Federated Learning presents a novel framework that adaptively selects the global learning rate based on both inter-client and coordinate-wise heterogeneity in the local updates. These advancements have the potential to significantly impact the field, enabling more efficient and private model training in a variety of applications.

Sources

Optimal Allocation of Privacy Budget on Hierarchical Data Release

FedDuA: Doubly Adaptive Federated Learning

Diffusion Learning with Partial Agent Participation and Local Updates

Heterogeneity-Aware Client Sampling: A Unified Solution for Consistent Federated Learning

Optimal Client Sampling in Federated Learning with Client-Level Heterogeneous Differential Privacy

Personalized Bayesian Federated Learning with Wasserstein Barycenter Aggregation

Simple and Optimal Algorithms for Heavy Hitters and Frequency Moments in Distributed Models

Federated prediction for scalable and privacy-preserved knowledge-based planning in radiotherapy

Privacy Preserving Conversion Modeling in Data Clean Room

Distributionally Robust Federated Learning with Client Drift Minimization

Optimal Piecewise-based Mechanism for Collecting Bounded Numerical Data under Local Differential Privacy

Federated Learning with Unlabeled Clients: Personalization Can Happen in Low Dimensions

Enhancing Federated Survival Analysis through Peer-Driven Client Reputation in Healthcare

Built with on top of