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

To leverage data and computation capabilities of mobile devices, machinelearning algorithms are deployed at the network edge for training artificialintelligence (AI) models, resulting in the new paradigm of edge learning. Inthis paper, we consider the framework of partitioned edge learning foriteratively training a large-scale model using many resource-constraineddevices (called workers). To this end, in each iteration, the model isdynamically partitioned into parametric blocks, which are downloaded to workergroups for updating using data subsets. Then, the local updates are uploaded toand cascaded by the server for updating a global model. To reduce resourceusage by minimizing the total learning-and-communication latency, this workfocuses on the novel joint design of parameter (computation load) allocationand bandwidth allocation (for downloading and uploading). Two design approachesare adopted. First, a practical sequential approach, called partiallyintegrated parameter-and-bandwidth allocation (PABA), yields two schemes,namely bandwidth aware parameter allocation and parameter aware bandwidthallocation. The former minimizes the load for the slowest (in computing) ofworker groups, each training a same parametric block. The latter allocates thelargest bandwidth to the worker being the latency bottleneck. Second, PABA arejointly optimized. Despite its being a nonconvex problem, an efficient andoptimal solution algorithm is derived by intelligently nesting a bisectionsearch and solving a convex problem. Experimental results using real datademonstrate that integrating PABA can substantially improve the performance ofpartitioned edge learning in terms of latency (by e.g., 46%) and accuracy (bye.g., 4%).

Joint Parameter-and-Bandwidth Allocation for Improving the Efficiency of Partitioned Edge Learning