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

Vision Transformers (ViT) and Visual Prompt Tuning (VPT) achievestate-of-the-art performance with improved efficiency in various computervision tasks. This suggests a promising paradigm shift of adapting pre-trainedViT models to Federated Learning (FL) settings. However, the challenge of dataheterogeneity among FL clients presents a significant hurdle in effectivelydeploying ViT models. Existing Generalized FL (GFL) and Personalized FL (PFL)methods have limitations in balancing performance across both global and localdata distributions. In this paper, we present a novel algorithm, SGPT, thatintegrates GFL and PFL approaches by employing a unique combination of bothshared and group-specific prompts. This design enables SGPT to capture bothcommon and group-specific features. A key feature of SGPT is its promptselection module, which facilitates the training of a single global modelcapable of automatically adapting to diverse local client data distributionswithout the need for local fine-tuning. To effectively train the prompts, weutilize block coordinate descent (BCD), learning from common featureinformation (shared prompts), and then more specialized knowledge (groupprompts) iteratively. Theoretically, we justify that learning the proposedprompts can reduce the gap between global and local performance. Empirically,we conduct experiments on both label and feature heterogeneity settings incomparison with state-of-the-art baselines, along with extensive ablationstudies, to substantiate the superior performance of SGPT.

Unlocking the Potential of Prompt-Tuning in Bridging Generalized and Personalized Federated Learning