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

Deep neural networks (DNNs) have achieved great success in various machinelearning tasks. However, most existing powerful DNN models are computationallyexpensive and memory demanding, hindering their deployment in devices with lowmemory and computational resources or in applications with strict latencyrequirements. Thus, several resource-adaptable or flexible approaches wererecently proposed that train at the same time a big model and severalresource-specific sub-models. Inplace knowledge distillation (IPKD) became apopular method to train those models and consists in distilling the knowledgefrom a larger model (teacher) to all other sub-models (students). In this worka novel generic training method called IPKD with teacher assistant (IPKD-TA) isintroduced, where sub-models themselves become teacher assistants teachingsmaller sub-models. We evaluated the proposed IPKD-TA training method using twostate-of-the-art flexible models (MSDNet and Slimmable MobileNet-V1) with twopopular image classification benchmarks (CIFAR-10 and CIFAR-100). Our resultsdemonstrate that the IPKD-TA is on par with the existing state of the art whileimproving it in most cases.

Inplace knowledge distillation with teacher assistant for improved training of flexible deep neural networks