Deep Learning With Edge Computing: A Review

Chen, Jiasi; Ran, Xukan

doi:10.1109/jproc.2019.2921977

Cited by 1,057 publications

(503 citation statements)

References 81 publications

Supporting

Mentioning

496

Contrasting

Unclassified

Order By: Relevance

“…Computer vision, Natural Language Processing (NLP), network functions, Internet-of-Things (IoT), and virtual/augmented reality applications utilising ML techniques have been developed for the edge computing context [8]. In order to meet computational challenges of the applications, collected data could be moved from end user devices back to the cloud for processing and ML [8]. However, latency, scalability and privacy challenges may need to be addressed for achieving satisfactory performance [8].…”

Section: Related Workmentioning

confidence: 99%

“…In order to meet computational challenges of the applications, collected data could be moved from end user devices back to the cloud for processing and ML [8]. However, latency, scalability and privacy challenges may need to be addressed for achieving satisfactory performance [8]. In the following, several approaches are described, in which latency of model inference has been optimised for the edge environment.…”

Section: Related Workmentioning

confidence: 99%

“…Sometimes models need to be designed or compressed for execution on resource-constrained devices [8]. For example, experiments with augmented reality enabling video processing framework [32] indicated that Raspberry Pi couldn't achieve a satisfying performance without optimisation, as only ~ 1 frame per second (FPS) could be processed.…”

Section: Related Workmentioning

confidence: 99%

“…Edge/5G computing is an emerging technological field [7], and the first products are being shipped to the markets. However, the utilisation of machine learning (ML) as part of the edge computing infrastructure is still an area for further research [8]. Particularly, it should be understood, how data is collected, and how models are Abstract Background: Augmented reality, computer vision and other (e.g.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Extending reference architecture of big data systems towards machine learning in edge computing environments

Pääkkönen

Pakkala

2020

J Big Data

View full text Add to dashboard Cite

IntroductionMany big data systems have been developed and realised to provide end user services (Netflix, Facebook, Twitter, LinkedIn etc.). Also, underlying architectures and technologies of the enabling systems have been published [1-3], and RAs have been designed and proposed [4][5][6]. Edge/5G computing is an emerging technological field [7], and the first products are being shipped to the markets. However, the utilisation of machine learning (ML) as part of the edge computing infrastructure is still an area for further research [8]. Particularly, it should be understood, how data is collected, and how models are Abstract Background: Augmented reality, computer vision and other (e.g. network functions, Internet-of-Things (IoT)) use cases can be realised in edge computing environments with machine learning (ML) techniques. For realisation of the use cases, it has to be understood how data is collected, stored, processed, analysed, and visualised in big data systems. In order to provide services with low latency for end users, often utilisation of ML techniques has to be optimized. Also, software/service developers have to understand, how to develop and deploy ML models in edge computing environments. Therefore, architecture design of big data systems to edge computing environments may be challenging. Findings:The contribution of this paper is reference architecture (RA) design of a big data system utilising ML techniques in edge computing environments. An earlier version of the RA has been extended based on 16 realised implementation architectures, which have been developed to edge/distributed computing environments. Also, deployment of architectural elements in different environments is described. Finally, a system view is provided of the software engineering aspects of ML model development and deployment. Conclusions:The presented RA may facilitate concrete architecture design of use cases in edge computing environments. The value of RAs is reduction of development and maintenance costs of systems, reduction of risks, and facilitation of communication between different stakeholders.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Extending reference architecture of big data systems towards machine learning in edge computing environments

Pääkkönen

Pakkala

2020

J Big Data

View full text Add to dashboard Cite

show abstract

“…Driven by this trend, it is urgent to push traditional cloud-based DNN models to the network edge so as to unleash the potentials of edge data and in turn provide intelligent services [4]- [5]. One possible architecture to perform inference tasks is on-device inference, i.e., running DNN models directly on MDs.…”

Section: Introductionmentioning

confidence: 99%

Reconfigurable Intelligent Surface for Green Edge Inference in Machine Learning

Song

Shi

2019

2019 IEEE Globecom Workshops (GC Wkshps)

View full text Add to dashboard Cite

Reconfigurable intelligent surface (RIS) as an emerging cost-effective technology can enhance the spectrum-and energy-efficiency of wireless networks. In this paper, we consider an RIS-aided green edge inference system, where the inference tasks generated from resource-limited mobile devices (MDs) are uploaded to and cooperatively performed at multiple resourceenhanced base stations (BSs). Taking into account both the computation and uplink/downlink transmit power consumption, we formulate an overall network power consumption minimization problem, which calls for the joint design of the set of tasks performed by each BS, transmit and receive beamforming vectors of the BSs, transmit power of the MDs, and uplink/downlink phase-shift matrices at the RIS. Such a problem is a mixed combinatorial optimization problem with nonconvex constraints and is highly intractable. To address the challenge of the combinatorial objective, a group sparse reformulation is proposed by exploiting the group sparsity structure of the beamforming vectors, while a block-structured optimization (BSO) approach is proposed to decouple the optimization variables. Finally, we propose a BSO with mixed 1,2-norm and difference-of-convex-functions (DC) based three-stage framework to solve the problem, where the mixed 1,2-norm is adopted to induce the group sparsity of beamforming vectors and DC is adopted to effectively handle the nonconvex rank-one constraint after matrix lifting. Numerical results demonstrate the supreme gain of deploying an RIS and confirm the effectiveness of the proposed algorithm over the baseline algorithms.Index Terms-Reconfigurable intelligent surface, joint uplink and downlink, green edge inference, block-structured optimization, mixed 1,2-norm,

show abstract

On the Pivotal Role of Artificial Intelligence Toward the Evolution of Smart Grids

Refaat¹,

Abu‐Rub²

2021

Smart Grid and Enabling Technologies

View full text Add to dashboard Cite

Deep Learning With Edge Computing: A Review

Cited by 1,057 publications

References 81 publications

Extending reference architecture of big data systems towards machine learning in edge computing environments

Extending reference architecture of big data systems towards machine learning in edge computing environments

Reconfigurable Intelligent Surface for Green Edge Inference in Machine Learning

On the Pivotal Role of Artificial Intelligence Toward the Evolution of Smart Grids

Contact Info

Product

Resources

About