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

Due to limited resources on edge and different characteristics of deep neuralnetwork (DNN) models, it is a big challenge to optimize DNN inferenceperformance in terms of energy consumption and end-to-end latency on edgedevices. In addition to the dynamic voltage frequency scaling (DVFS) technique,the edge-cloud architecture provides a collaborative approach for efficient DNNinference. However, current edge-cloud collaborative inference methods have notoptimized various compute resources on edge devices. Thus, we propose DVFO, anovel DVFS-enabled edge-cloud collaborative inference framework, whichco-optimizes DVFS and offloading parameters via deep reinforcement learning(DRL). Specifically, DVFO automatically co-optimizes 1) the CPU, GPU and memoryfrequencies of edge devices, and 2) the feature maps to be offloaded to cloudservers. In addition, it leverages a thinking-while-moving concurrent mechanismto accelerate the DRL learning process, and a spatial-channel attentionmechanism to extract DNN feature maps of secondary importance for workloadoffloading. This approach improves inference performance for different DNNmodels under various edge-cloud network conditions. Extensive evaluations usingtwo datasets and six widely-deployed DNN models on three heterogeneous edgedevices show that DVFO significantly reduces the energy consumption by 33% onaverage, compared to state-of-the-art schemes. Moreover, DVFO achieves up to28.6%-59.1% end-to-end latency reduction, while maintaining accuracy within 1%loss on average.

DVFO: Learning-Based DVFS for Energy-Efficient Edge-Cloud Collaborative Inference