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

Deep Neural Networks (DNNs) have been ubiquitously adopted in internet ofthings and are becoming an integral of our daily life. When tackling theevolving learning tasks in real world, such as classifying different types ofobjects, DNNs face the challenge to continually retrain themselves according tothe tasks on different edge devices. Federated continual learning is apromising technique that offers partial solutions but yet to overcome thefollowing difficulties: the significant accuracy loss due to the limitedon-device processing, the negative knowledge transfer caused by the limitedcommunication of non-IID data, and the limited scalability on the tasks andedge devices. In this paper, we propose FedKNOW, an accurate and scalablefederated continual learning framework, via a novel concept of signature taskknowledge. FedKNOW is a client side solution that continuously extracts andintegrates the knowledge of signature tasks which are highly influenced by thecurrent task. Each client of FedKNOW is composed of a knowledge extractor, agradient restorer and, most importantly, a gradient integrator. Upon trainingfor a new task, the gradient integrator ensures the prevention of catastrophicforgetting and mitigation of negative knowledge transfer by effectivelycombining signature tasks identified from the past local tasks and otherclients' current tasks through the global model. We implement FedKNOW inPyTorch and extensively evaluate it against state-of-the-art techniques usingpopular federated continual learning benchmarks. Extensive evaluation resultson heterogeneous edge devices show that FedKNOW improves model accuracy by63.24% without increasing model training time, reduces communication cost by34.28%, and achieves more improvements under difficult scenarios such as largenumbers of tasks or clients, and training different complex networks.

FedKNOW: Federated Continual Learning with Signature Task Knowledge Integration at Edge