MUFFLe: Efficient Model Update Compression via Generalized Deduplication for Federated Learning
arXiv:2606.14354v1 Announce Type: new Abstract: Federated learning is well suited to edge environments but is often limited by the uplink cost of transmitting model updates. This Work-in-Progress paper presents MUFFLe, a communication-efficient update compression scheme that integrates generalized deduplication (GD) into the FedAvg pipeline. MUFFLe deduplicates repeated patterns across the update vector, yielding a fixed-rate, variable-count compression scheme. Preliminary experiments on IID MNIST with 20 clients show that MUFFLe reaches the target accuracy of $92.93\%$ with 38~MB cumulative u
The accelerating deployment of federated learning in edge environments, particularly with the proliferation of AI-enabled devices, is driving intense demand for efficient model update compression.
Efficient communication is a critical bottleneck for scaling federated learning, impacting everything from device battery life to network bandwidth, and directly influencing the feasibility of widespread FL applications.
This advancement provides a tangible method to significantly reduce uplink costs in federated learning, making it more practical for resource-constrained edge devices and potentially accelerating adoption.
- · Edge device manufacturers
- · Federated learning platforms
- · Telecommunication companies
- · AI developers
- · Companies relying on outdated FL communication protocols
Reduced data transmission costs and improved efficiency for federated learning deployments.
Faster and more scalable development of AI models on distributed data sources without centralizing sensitive information.
Enhanced privacy-preserving AI applications and a broader penetration of AI into highly distributed and resource-limited environments.
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