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

Data heterogeneity presents significant challenges for federated learning(FL). Recently, dataset distillation techniques have been introduced, andperformed at the client level, to attempt to mitigate some of these challenges.In this paper, we propose a highly efficient FL dataset distillation frameworkon the server side, significantly reducing both the computational andcommunication demands on local devices while enhancing the clients' privacy.Unlike previous strategies that perform dataset distillation on local devicesand upload synthetic data to the server, our technique enables the server toleverage prior knowledge from pre-trained deep generative models to synthesizeessential data representations from a heterogeneous model architecture. Thisprocess allows local devices to train smaller surrogate models while enablingthe training of a larger global model on the server, effectively minimizingresource utilization. We substantiate our claim with a theoretical analysis,demonstrating the asymptotic resemblance of the process to the hypotheticalideal of completely centralized training on a heterogeneous dataset. Empiricalevidence from our comprehensive experiments indicates our method's superiority,delivering an accuracy enhancement of up to 40% over non-dataset-distillationtechniques in highly heterogeneous FL contexts, and surpassing existingdataset-distillation methods by 18%. In addition to the high accuracy, ourframework converges faster than the baselines because rather than the servertrains on several sets of heterogeneous data distributions, it trains on amulti-modal distribution. Our code is available athttps://github.com/FedDG23/FedDG-main.git

Unlocking the Potential of Federated Learning: The Symphony of Dataset Distillation via Deep Generative Latents