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 (FL) is an emerging machine learning paradigm in which acentral server coordinates multiple participants (a.k.a. FL clients) to train amodel collaboratively on decentralized data with privacy protection. Thisparadigm constrains that all clients have to train models with the samestructures (homogeneous). In practice, FL often faces statisticalheterogeneity, system heterogeneity and model heterogeneity challenges. Thesechallenging issues inspire the field of Model-Heterogeneous PersonalizedFederated Learning (MHPFL) which aims to train a personalized and heterogeneouslocal model for each FL client. Existing MHPFL approaches cannot achievesatisfactory model performance, acceptable computational overhead and efficientcommunication simultaneously. To bridge this gap, we propose a novelcomputation- and communication-efficient model-heterogeneous personalizedFederated learning framework based on LoRA tuning (FedLoRA). It is designed toincorporate a homogeneous small adapter for each client's heterogeneous localmodel. Both models are trained following the proposed iterative training forglobal-local knowledge exchange. The homogeneous small local adapters are sentto the FL server to be aggregated into a global adapter. In this way, FLclients can train heterogeneous local models without incurring high computationand communication costs. We theoretically prove the non-convex convergence rateof FedLoRA. Extensive experiments on two real-world datasets demonstrate thatFedLoRA outperforms six state-of-the-art baselines, beating the best approachby 1.35% in terms of test accuracy, 11.81 times computation overhead reductionand 7.41 times communication cost saving.

FedLoRA: Model-Heterogeneous Personalized Federated Learning with LoRA Tuning