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

The pervasiveness of "Internet-of-Things" in our daily life has led to arecent surge in fog computing, encompassing a collaboration of cloud computingand edge intelligence. To that effect, deep learning has been a major drivingforce towards enabling such intelligent systems. However, growing model sizesin deep learning pose a significant challenge towards deployment inresource-constrained edge devices. Moreover, in a distributed intelligenceenvironment, efficient workload distribution is necessary between edge andcloud systems. To address these challenges, we propose a conditionally deephybrid neural network for enabling AI-based fog computing. The proposed networkcan be deployed in a distributed manner, consisting of quantized layers andearly exits at the edge and full-precision layers on the cloud. Duringinference, if an early exit has high confidence in the classification results,it would allow samples to exit at the edge, and the deeper layers on the cloudare activated conditionally, which can lead to improved energy efficiency andinference latency. We perform an extensive design space exploration with thegoal of minimizing energy consumption at the edge while achievingstate-of-the-art classification accuracies on image classification tasks. Weshow that with binarized layers at the edge, the proposed conditional hybridnetwork can process 65% of inferences at the edge, leading to 5.5xcomputational energy reduction with minimal accuracy degradation on CIFAR-10dataset. For the more complex dataset CIFAR-100, we observe that the proposednetwork with 4-bit quantization at the edge achieves 52% early classificationat the edge with 4.8x energy reduction. The analysis gives us insights ondesigning efficient hybrid networks which achieve significantly higher energyefficiency than full-precision networks for edge-cloud based distributedintelligence systems.

Conditionally Deep Hybrid Neural Networks Across Edge and Cloud