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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Federated Learning (FL) is an emerging approach for collaboratively trainingDeep Neural Networks (DNNs) on mobile devices, without private user dataleaving the devices. Previous works have shown that non-Independent andIdentically Distributed (non-IID) user data harms the convergence speed of theFL algorithms. Furthermore, most existing work on FL measures global-modelaccuracy, but in many cases, such as user content-recommendation, improvingindividual User model Accuracy (UA) is the real objective. To address theseissues, we propose a Multi-Task FL (MTFL) algorithm that introducesnon-federated Batch-Normalization (BN) layers into the federated DNN. MTFLbenefits UA and convergence speed by allowing users to train modelspersonalised to their own data. MTFL is compatible with popular iterative FLoptimisation algorithms such as Federated Averaging (FedAvg), and we showempirically that a distributed form of Adam optimisation (FedAvg-Adam) benefitsconvergence speed even further when used as the optimisation strategy withinMTFL. Experiments using MNIST and CIFAR10 demonstrate that MTFL is able tosignificantly reduce the number of rounds required to reach a target UA, by upto $5\times$ when using existing FL optimisation strategies, and with a further$3\times$ improvement when using FedAvg-Adam. We compare MTFL to competingpersonalised FL algorithms, showing that it is able to achieve the best UA forMNIST and CIFAR10 in all considered scenarios. Finally, we evaluate MTFL withFedAvg-Adam on an edge-computing testbed, showing that its convergence and UAbenefits outweigh its overhead.

Multi-Task Federated Learning for Personalised Deep Neural Networks in Edge Computing