Actions at the Edge: Jointly Optimizing the Resources in Multi-Access Edge Computing

Deng, Yanhua; Zhu, Guangyu; Fang, Yuguang; Chen, Zhigang; Deng, Xiaoheng

doi:10.1109/mwc.006.2100699

Cited by 22 publications

(8 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Then, EPSL performs last-layer activations' gradient aggregation on the dimension of client devices. In other words, each client device employs ⌈φb⌉ out of b of its lastlayer activations' gradients for aggregation with other client devices, after which the aggregated activations' gradients go 2 The server can execute FP process for multiple client devices in either a serial or parallel fashion, where the latency in (16) would not be affected. µ j (̟ ℓc+ℓs−1 − ̟ j ) denote the computation workload of the server-side BP process for each data sample (excluding the last layer), where ̟ j is the BP computation workload of propagating the first j layer neural network for one data sample.…”

Section: A Training Latency Computationmentioning

confidence: 99%

“…Although training models for a reasonable number of client devices might not be a crucial issue for a sufficiently powerful cloud computing center, it is arguably overwhelming to a resource-constrained edge server as the number of served client devices increases. Particularly, edge computing servers in 5G and beyond can be (small) base stations and access points usually equipped with limited capabilities [15,16]. On the other hand, communication latency is a limiting factor due to the large volume of cutlayer data and model exchange involved in SL.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Efficient Parallel Split Learning over Resource-constrained Wireless Edge Networks

Zheng¹,

Zhu²,

Deng³

et al. 2023

Preprint

View full text Add to dashboard Cite

The increasingly deeper neural networks hinder the democratization of privacy-enhancing distributed learning, such as federated learning (FL), to resource-constrained devices. To overcome this challenge, in this paper, we advocate the integration of edge computing paradigm and parallel split learning (PSL), allowing multiple client devices to offload substantial training workloads to an edge server via layer-wise model split. By observing that existing PSL schemes incur excessive training latency and large volume of data transmissions, we propose an innovative PSL framework, namely, efficient parallel split learning (EPSL), to accelerate model training. To be specific, EPSL parallelizes client-side model training and reduces the dimension of activations' gradients for back propagation (BP) via last-layer gradient aggregation, leading to a significant reduction in server-side training and communication latency. Moreover, by considering the heterogeneous channel conditions and computing capabilities at client devices, we jointly optimize subchannel allocation, power control, and cut layer selection to minimize the per-round latency. Simulation results show that the proposed EPSL framework significantly decreases the training latency needed to achieve a target accuracy compared with the stateof-the-art benchmarks, and the tailored resource management and layer split strategy can considerably reduce latency than the counterpart without optimization.

show abstract

Section: A Training Latency Computationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Efficient Parallel Split Learning over Resource-constrained Wireless Edge Networks

Zheng¹,

Zhu²,

Deng³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Minimizing the number of update rounds, efficient spectral usage, and cost-effective communication are a few challenges of federated learning. Multiaccess edge computing (MEC) is an emerging technology [423] that has the potential to ensure optimal coordination between the spectral, storage, and computing resources within the limited power and latency budget. By sharing the communication resources in proximity, MEC has the potential to meet the spectral demand in data collection and distributed training and inference.…”

Section: K Increased Demand Of Communication Resourcesmentioning

confidence: 99%

Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review

et al. 2023

View full text Add to dashboard Cite

Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly accurate models for data-driven, learned, automatic, and practical machine learning (ML) solutions to end-user applications remains challenging. DL algorithms are often computationally expensive, power-hungry, and require large memory to process complex and iterative operations of millions of parameters. Hence, training and inference of DL models are typically performed on high-performance computing (HPC) clusters in the cloud. Data transmission to the cloud results in high latency, round-trip delay, security and privacy concerns, and the inability of real-time decisions. Thus, processing on edge devices can significantly reduce cloud transmission cost. Edge devices are end devices closest to the user, such as mobile phones, cyber-physical systems (CPSs), wearables, the Internet of Things (IoT), embedded and autonomous systems, and

show abstract

“…Multi-access edge computing (MEC) has been identified as a promising architecture for computing services that aims to provide real-time or low latency services to end-users located in close proximity [1], [2]. One of the primary techniques utilized in MEC is computation offloading, which enables computing tasks to be processed locally or offloaded to an edge server (ES) based on the availability of local computing resources and transmission conditions [3].…”

Section: Introductionmentioning

confidence: 99%

Task Offloading in Multi-Hop Relay-Aided Multi-Access Edge Computing

Deng

Chen

Fang

2023

IEEE Trans. Veh. Technol.

Self Cite

View full text Add to dashboard Cite

Multi-access edge computing (MEC) is a promising technology to enhance the quality of service, particularly for low-latency services, by enabling computing offloading to edge servers (ESs) in close proximity. To avoid network congestion, collaborative edge computing has become an emerging paradigm to enable different ESs to collaboratively share their data and computation resources. However, most papers in collaborative edge computing only allow one-hop offloading, which may limit computing resource sharing due to either poor channel conditions or computing workload at ESs one-hop away. By allowing ESs multi-hop away to also share the computing workload, a multihop MEC enables more ESs to share their computing resources. Inspired by this observation, in this paper, we propose to leverage omnipresent vehicles in a city to form a data transportation network for task delivery in a multi-hop fashion. Here, we propose a general multi-hop task offloading framework for vehicle-assisted MEC where tasks from users can be offloaded to powerful ESs via potentially multi-hop transmissions. Under the proposed framework, we develop a reinforcement learning based task offloading approach to address the curse of dimensionality problem due to vehicular mobility and channel variability, with the goal to maximize the aggregated service throughput under constraints on end-to-end latency, spectrum, and computing resources. Numerical results demonstrate that the proposed algorithm achieves excellent performance with low complexity and outperforms existing benchmark schemes.

show abstract

Actions at the Edge: Jointly Optimizing the Resources in Multi-Access Edge Computing

Cited by 22 publications

References 12 publications

Efficient Parallel Split Learning over Resource-constrained Wireless Edge Networks

Efficient Parallel Split Learning over Resource-constrained Wireless Edge Networks

Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review

Task Offloading in Multi-Hop Relay-Aided Multi-Access Edge Computing

Contact Info

Product

Resources

About