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

Learning a global model by abstracting the knowledge, distributed acrossmultiple clients, without aggregating the raw data is the primary goal ofFederated Learning (FL). Typically, this works in rounds alternating betweenparallel local training at several clients, followed by model aggregation at aserver. We found that existing FL methods under-perform when local datasets aresmall and present severe label skew as these lead to over-fitting and localmodel bias. This is a realistic setting in many real-world applications. Toaddress the problem, we propose \textit{FLea}, a unified framework that tacklesover-fitting and local bias by encouraging clients to exchangeprivacy-protected features to aid local training. The features refer toactivations from an intermediate layer of the model, which are obfuscatedbefore being shared with other clients to protect sensitive information in thedata. \textit{FLea} leverages a novel way of combining local and sharedfeatures as augmentations to enhance local model learning. Our extensiveexperiments demonstrate that \textit{FLea} outperforms the start-of-the-art FLmethods, sharing only model parameters, by up to $17.6\%$, and FL methods thatshare data augmentations by up to $6.3\%$, while reducing the privacyvulnerability associated with shared data augmentations.

FLea: Improving federated learning on scarce and label-skewed data via privacy-preserving feature augmentation