An Integrated Cloud-Edge-Device Adaptive Deep Learning Service for Cross-Platform Web

Yakun Huang, Xiuquan Qiao, Jian Tang, Pei Ren, Ling Liu, Calton Pu, Jun Liang Chen

Research output: Contribution to journalArticlepeer-review

Abstract

Deep learning shows great promise in providing more intelligence to the cross-platform web, but insufficient infrastructure, heavy models, and intensive computation limit the use of deep learning with low-performing web inference libraries. This paper proposes DeepAdapter, an integrated cloud-edge-device framework that ties the edge, the remote cloud, with the device by cross-platform web technology for adaptive deep learning services towards lower processing latency, lower mobile energy, and higher system throughput. First, we provide a context-aware pruning algorithm that incorporates the latency, the network condition, and the device's computing capability to fit the cross-platform web's resource constraints better. Second, DeepAdapter provides a model cache update mechanism improving the model request hit rate for cross-platform web users. Third, DeepAdapter provides an online matcher for dispatching appropriate models to service requestors and employs a collaborative mechanism to ensure accuracy. Last, DeepScheduler in DeepAdapter plays a significant role in scheduling high concurrent requests among various edge centers by designing the reward prediction model that improves the DRL training. Extensive experiments show that DeepAdapter can decrease average latency by 1.33x, reduce average mobile energy by 1.4x, and improve system throughput by 2.1x with considerable accuracy.

Original languageEnglish (US)
JournalIEEE Transactions on Mobile Computing
DOIs
StateAccepted/In press - 2021
Externally publishedYes

Keywords

  • Adaptation models
  • adaptive service
  • Cloud computing
  • Collaboration
  • Computational modeling
  • Context modeling
  • cross-platform web
  • Deep learning
  • DNN compression
  • edge computing
  • Load modeling
  • Mobile computing

ASJC Scopus subject areas

  • Software
  • Computer Networks and Communications
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'An Integrated Cloud-Edge-Device Adaptive Deep Learning Service for Cross-Platform Web'. Together they form a unique fingerprint.

Cite this