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

Federated learning alleviates the privacy risk in distributed learning bytransmitting only the local model updates to the central server. However, itfaces challenges including statistical heterogeneity of clients' datasets andresource constraints of client devices, which severely impact the trainingperformance and user experience. Prior works have tackled these challenges bycombining personalization with model compression schemes including quantizationand pruning. However, the pruning is data-dependent and thus must be done onthe client side which requires considerable computation cost. Moreover, thepruning normally trains a binary supermask $\in \{0, 1\}$ which significantlylimits the model capacity yet with no computation benefit. Consequently, thetraining requires high computation cost and a long time to converge while themodel performance does not pay off. In this work, we propose HideNseek whichemploys one-shot data-agnostic pruning at initialization to get a subnetworkbased on weights' synaptic saliency. Each client then optimizes a signsupermask $\in \{-1, +1\}$ multiplied by the unpruned weights to allow fasterconvergence with the same compression rates as state-of-the-art. Empiricalresults from three datasets demonstrate that compared to state-of-the-art,HideNseek improves inferences accuracies by up to 40.6\% while reducing thecommunication cost and training time by up to 39.7\% and 46.8\% respectively.

HideNseek: Federated Lottery Ticket via Server-side Pruning and Sign Supermask