HierTrain: Fast Hierarchical Edge AI Learning With Hybrid Parallelism in Mobile-Edge-Cloud Computing

Liu, Deyin; Chen, Xu; Zhou, Zhi; Ling, Qing

doi:10.1109/ojcoms.2020.2994737

Cited by 43 publications

(23 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our agent first explores 100 episodes with constant noise σ � 0.5 and then explores 300 episodes with exponential attenuation noise σ, with an attenuation coefficient of 0.99. We set the cloud-based computation offloading method [49] as the baseline and added the status quo method HierTrain [50] as a comparison. We compared the performance of the Cogent architecture with the baseline and status quo in terms of inference latency (Section 4.2).…”

Section: Methodsmentioning

confidence: 99%

Collaborative Intelligence: Accelerating Deep Neural Network Inference via Device-Edge Synergy

Shan

Cui

2020

Security and Communication Networks

View full text Add to dashboard Cite

With the development of mobile edge computing (MEC), more and more intelligent services and applications based on deep neural networks are deployed on mobile devices to meet the diverse and personalized needs of users. Unfortunately, deploying and inferencing deep learning models on resource-constrained devices are challenging. The traditional cloud-based method usually runs the deep learning model on the cloud server. Since a large amount of input data needs to be transmitted to the server through WAN, it will cause a large service latency. This is unacceptable for most current latency-sensitive and computation-intensive applications. In this paper, we propose Cogent, an execution framework that accelerates deep neural network inference through device-edge synergy. In the Cogent framework, it is divided into two operation stages, including the automatic pruning and partition stage and the containerized deployment stage. Cogent uses reinforcement learning (RL) to automatically predict pruning and partition strategies based on feedback from the hardware configuration and system conditions so that the pruned and partitioned model can better adapt to the system environment and user hardware configuration. Then through containerized deployment to the device and the edge server to accelerate model inference, experiments show that the learning-based hardware-aware automatic pruning and partition scheme can significantly reduce the service latency, and it accelerates the overall model inference process while maintaining accuracy. Using this method can accelerate up to 8.89× without loss of accuracy of more than 7%.

show abstract

Section: Methodsmentioning

confidence: 99%

Collaborative Intelligence: Accelerating Deep Neural Network Inference via Device-Edge Synergy

Shan

Cui

2020

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…[6] discussed the research achievements from different perspectives combining edge cloud collaboration and edge intelligence with edge computing security. In [7] a hardware prototype was implemented to provide extensive experiments on a deep neural networks (DNNs) to achieve minimum training time. [8] proposed a model for running application monitoring in an Internet of Thing (IoT) environment.…”

Section: Related Workmentioning

confidence: 99%

“…While the power in the battery b i (k) is limited to the maximum battery capacity B max i as in (7):…”

Section: B Proposed System Model Parametersmentioning

confidence: 99%

Orchestration of MEC Computation Jobs and Energy Consumption Challenges in 5G and Beyond

2022

View full text Add to dashboard Cite

Mobile Edge Computing (MEC) technology philosophy inspires the next generation mobile networks to provide cloud computing capabilities in addition to a diverse range of Information Technology (IT) services with ultra-low latency and higher bandwidth at the edge. One of the most common challenges of 5G-MEC is the management and orchestration across all networks and infrastructure resources as well as end-to-end quality of experience. The decentralized architecture of MEC with independent and noncollaborative servers results in the situation of having underutilized servers with wasted energy. Moreover, the consequences of having highly utilized servers with highly consumed energy are not only the incapability to accommodate all the load of the computing jobs and the dramatic increase in the total OPEX cost, but it also creates some environmental problems. Orchestrating servers' workload and control offloading the computation jobs is one of the technical advantages of MEC since it satisfies the increasing requirements of modern mobile applications while optimizing the energy consumption and cost. In this work, we consider cluster-based energy-aware offloading framework. The proposed work consists of dual-tier domain divided into clusters of Edge Servers ES s . We have presented the results of our simulation as a proof of our concept that the formulated adaptive strategy to minimize the optimization problem calculation per cluster reduces the energy consumption and enhances the quality of experience while achieving the conservation of the related computing and storage resources cost.

show abstract

“…For intelligent IoT applications, a framework is proposed by Liu et al [20], which is based on the cloud edge architecture, apply federal learning to make smart applications available. In order to solve the heterogeneous problem in the IoT environment.…”

Section: Related Workmentioning

confidence: 99%

Collaborative Edge Computing and Caching With Deep Reinforcement Learning Decision Agents

et al. 2020

View full text Add to dashboard Cite

Large amounts of data will be generated due to the rapid development of the Internet of Things (IoT) technologies and 5th generation mobile networks (5G), the processing and analysis requirements of big data will challenge existing networks and processing platforms. As the most promising technology in 5G networks, edge computing will greatly ease the pressure on network and data processing analysis on the edge. In this paper, we considered the coordination between compute and cache resources between multi-level edge computing nodes (ENs), users under this system can offload computing tasks to ENs to improve quality of service (QoS). We aimed to maximize the long-term profit on the edge, while satisfying the low-latency computing of the users, and jointly optimize the edge-side node offloading strategy and resource allocation. However, it is challenging to obtain an optimal strategy in such a dynamic and complex system. To solve the complex resource allocation problem on the edge and make edge have certain adaptation and cooperation, we used double deep Q-learning (DDQN) to make decisions, ability to maximize long-term gains while making quick decisions. The simulation results prove the effectiveness of DDQN in maximizing revenue when allocation resources on the edge.

show abstract

HierTrain: Fast Hierarchical Edge AI Learning With Hybrid Parallelism in Mobile-Edge-Cloud Computing

Cited by 43 publications

References 28 publications

Collaborative Intelligence: Accelerating Deep Neural Network Inference via Device-Edge Synergy

Collaborative Intelligence: Accelerating Deep Neural Network Inference via Device-Edge Synergy

Orchestration of MEC Computation Jobs and Energy Consumption Challenges in 5G and Beyond

Collaborative Edge Computing and Caching With Deep Reinforcement Learning Decision Agents

Contact Info

Product

Resources

About