Communication-Efficient Edge AI: Algorithms and Systems

Shi, Yuanming; Yang, Kai; Jiang, Tao; Zhang, Jun; Letaief, Khaled B.

doi:10.1109/comst.2020.3007787

Cited by 329 publications

(164 citation statements)

References 126 publications

Supporting

Mentioning

164

Contrasting

Order By: Relevance

“…In the last years, we have seen the rise of powerful edge devices that provide the computational power to run AI algorithms at the edge, near to the data sources and in realtime [59]. Since algorithms running on the edge can provide a fast response to condition change and also reduce the size of data to be sent to the cloud by implementing preprocessing methods, and AI algorithms running in the cloud can provide optimization, prediction, and planning functions.…”

Section: Hpc and Processing Infrastructuresmentioning

confidence: 99%

Quo Vadis Industry 4.0? Position, Trends, and Challenges

Leitão

Pires

Karnouskos

et al. 2020

IEEE Open J. Ind. Electron. Soc.

View full text Add to dashboard Cite

Industry 4.0 vision and its mandated digital transformation are radically reshaping the way business is carried out and the way overall industrial processes and collaborations are operating. In this work, the objective is to analyze the current level of adoption of Industry 4.0, via the footprint available in industrial and academic research works. The analysis performed reveals insights on how Industry 4.0 has impacted and is still influencing research and innovation in industrial systems, services, and business approaches. It also reveals pertinent trends on key enabling features, technologies and challenges associated with this 4 th industrial revolution, mainly focusing on the pathways for wider industrial adoption of Industry 4.0-compliant technologies and solutions.

show abstract

Section: Hpc and Processing Infrastructuresmentioning

confidence: 99%

Quo Vadis Industry 4.0? Position, Trends, and Challenges

Leitão

Pires

Karnouskos

et al. 2020

IEEE Open J. Ind. Electron. Soc.

View full text Add to dashboard Cite

show abstract

“…Currently, model inferences are mostly performed in the cloud; but as the diversity of DNN applications grows, other alternatives to the centralized training and inference strategy are required to lessen the burden on network infrastructures [12]. To that, we opt for Edge Computing [19]: a distributed computing paradigm where software-defined networks are built to decentralize data and provide results expected to be the same as which of cloud computing [20]. Solving these problems, edge DNN aims to process DNN models directly on edge devices.…”

Section: Related Work 21 Edge-based Vs Cloud-based Aimentioning

confidence: 99%

“…Solving these problems, edge DNN aims to process DNN models directly on edge devices. However, to utilize their innate computing power, edge computing faces more resource allocation problems, due to the inherent difference in hardware architecture, the need to sample inputs from built-in peripherals, bandwidth constraints and more [18,20]. To offset these problems, deployment on edge-devices requires DNN models to be minimal enough for fast inference and low-latency updates.…”

Section: Related Work 21 Edge-based Vs Cloud-based Aimentioning

confidence: 99%

Efficient Low-Latency Dynamic Licensing for Deep Neural Network Deployment on Edge Devices

Van¹,

Tran²,

Minh³

et al. 2020

Proceedings of the 2020 3rd International Conference on Computational Intelligence and Intelligent Systems

View full text Add to dashboard Cite

Along with the rapid development in the field of artificial intelligence (AI), especially deep learning, deep neural network (DNN) applications are becoming more and more popular in reality. To be able to withstand the heavy load from mainstream users, deployment techniques are essential in bringing neural network models from research to production. Among the two popular computing topologies for deploying neural network models in production are cloud-computing and edge-computing. Recent advances in communication technologies, along with the great increase in the number of mobile devices, has made edge-computing gradually become an inevitable trend. In this paper, we propose an architecture to solve deploying and processing deep neural networks on edge-devices by leveraging their synergy with the cloud and the access-control mechanisms of the database. Adopting this architecture allows lowlatency DNN model updates on devices. At the same time, with only one model deployed, we can easily make different versions of it by setting access permissions on the model weights. This method allows for dynamic model licensing, which benefits commercial applications. CCS CONCEPTS • Software and its engineering → Software version control; • Computer systems organization → Embedded software.

show abstract

“…In contrast, the edge-only inference incurs excessive communication overhead caused by the large volume of input data. Fortunately, the device-edge co-inference paradigm [10,11], which forwards an intermediate feature to the edge server for processing, is a promising candidate to reduce the inference latency by striking a balance between the computation and communication overhead.…”

Section: Introductionmentioning

confidence: 99%

Branchy-GNN: A Device-Edge Co-Inference Framework for Efficient Point Cloud Processing

Shao

Zhang

Mao

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

The recent advancements of three-dimensional (3D) data acquisition devices have spurred a new breed of applications that rely on point cloud data processing. However, processing a large volume of point cloud data brings a significant workload on resource-constrained mobile devices, prohibiting from unleashing their full potentials. Built upon the emerging paradigm of device-edge co-inference, where an edge device extracts and transmits the intermediate feature to an edge server for further processing, we propose Branchy-GNN for efficient graph neural network (GNN) based point cloud processing by leveraging edge computing platforms. In order to reduce the on-device computational cost, the Branchy-GNN adds branch networks for early exiting. Besides, it employs learning-based joint source-channel coding (JSCC) for the intermediate feature compression to reduce the communication overhead. Our experimental results demonstrate that the proposed Branchy-GNN secures a significant latency reduction compared with several benchmark methods.

show abstract

Communication-Efficient Edge AI: Algorithms and Systems

Cited by 329 publications

References 126 publications

Quo Vadis Industry 4.0? Position, Trends, and Challenges

Quo Vadis Industry 4.0? Position, Trends, and Challenges

Efficient Low-Latency Dynamic Licensing for Deep Neural Network Deployment on Edge Devices

Branchy-GNN: A Device-Edge Co-Inference Framework for Efficient Point Cloud Processing

Contact Info

Product

Resources

About