English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

To enable large-scale machine learning in bandwidth-hungry environments suchas wireless networks, significant progress has been made recently in designingcommunication-efficient federated learning algorithms with the aid ofcommunication compression. On the other end, privacy-preserving, especially atthe client level, is another important desideratum that has not been addressedsimultaneously in the presence of advanced communication compression techniquesyet. In this paper, we propose a unified framework that enhances thecommunication efficiency of private federated learning with communicationcompression. Exploiting both general compression operators and localdifferential privacy, we first examine a simple algorithm that appliescompression directly to differentially-private stochastic gradient descent, andidentify its limitations. We then propose a unified framework SoteriaFL forprivate federated learning, which accommodates a general family of localgradient estimators including popular stochastic variance-reduced gradientmethods and the state-of-the-art shifted compression scheme. We provide acomprehensive characterization of its performance trade-offs in terms ofprivacy, utility, and communication complexity, where SoteraFL is shown toachieve better communication complexity without sacrificing privacy nor utilitythan other private federated learning algorithms without communicationcompression.

SoteriaFL: A Unified Framework for Private Federated Learning with Communication Compression