Advances in Federated Learning and Machine Unlearning

The field of federated learning and machine unlearning is rapidly evolving, with a focus on developing innovative methods to protect data privacy and ensure efficient model training. Recent research has explored the use of novel frameworks, such as PURGE, DELETE, and FedPaI, to achieve efficient machine unlearning and extreme sparsity in federated learning. These approaches have shown significant improvements in terms of accuracy, communication efficiency, and computational overhead. Additionally, the use of edge model overlays, selective pruning, and knowledge deletion techniques has been investigated to further enhance the performance and privacy of federated learning models. Noteworthy papers include the proposal of the PURGE framework for efficient verified machine unlearning, the introduction of the DELETE method for general unlearning in class-centric tasks, and the development of the FedPaI framework for achieving extreme sparsity in federated learning. These advancements have the potential to enable more efficient and private federated learning, and are expected to have a significant impact on the field in the coming years.

Sources

Efficient Verified Machine Unlearning For Distillation

Buyer-Initiated Auction Mechanism for Data Redemption in Machine Unlearning

Decoupled Distillation to Erase: A General Unlearning Method for Any Class-centric Tasks

Federated Structured Sparse PCA for Anomaly Detection in IoT Networks

FedPaI: Achieving Extreme Sparsity in Federated Learning via Pruning at Initialization

Benchmarking Federated Machine Unlearning methods for Tabular Data

EMO: Edge Model Overlays to Scale Model Size in Federated Learning

A Two-Timescale Approach for Wireless Federated Learning with Parameter Freezing and Power Control

Split Federated Learning for UAV-Enabled Integrated Sensing, Computation, and Communication

Sky of Unlearning (SoUL): Rewiring Federated Machine Unlearning via Selective Pruning

ESC: Erasing Space Concept for Knowledge Deletion

Tree-based Models for Vertical Federated Learning: A Survey

Built with on top of