Energy-Efficient Machine Learning on the Edges

Kumar, Mohit; Zhang, Xingzhou; Liu, Liangkai; Wang, Yifan; Shi, Wei

doi:10.1109/ipdpsw50202.2020.00153

Cited by 23 publications

(18 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed method is low-power, small area and high resolution DPWM design that prepares for DC-DC converter to power the ULV operating IoT networks. Energy-efficient machine learning on the edges for IoT devices were carried out by [160]. The proposed algorithm has shown up to 14.46% improvement in energy consumption.…”

Section: Aspects Of Energy-efficiency Optimization Methodsmentioning

confidence: 99%

Energy Efficiency for Green Internet of Things (IoT) Networks: A Survey

Farhan

Hameed

Ahmed

et al. 2021

Network

View full text Add to dashboard Cite

The last decade has witnessed the rise of the proliferation of Internet-enabled devices. The Internet of Things (IoT) is becoming ever more pervasive in everyday life, connecting an ever-greater array of diverse physical objects. The key vision of the IoT is to bring a massive number of smart devices together in integrated and interconnected heterogeneous networks, making the Internet even more useful. Therefore, this paper introduces a brief introduction to the history and evolution of the Internet. Then, it presents the IoT, which is followed by a list of application domains and enabling technologies. The wireless sensor network (WSN) is revealed as one of the important elements in IoT applications, and the paper describes the relationship between WSNs and the IoT. This research is concerned with developing energy-efficiency techniques for WSNs that enable the IoT. After having identified sources of energy wastage, this paper reviews the literature that discusses the most relevant methods to minimizing the energy exhaustion of IoT and WSNs. We also identify the gaps in the existing literature in terms of energy preservation measures that could be researched and it can be considered in future works. The survey gives a near-complete and up-to-date view of the IoT in the energy field. It provides a summary and recommendations of a large range of energy-efficiency methods proposed in the literature that will help and support future researchers. Please note that the manuscript is an extended version and based on the summary of the Ph.D. thesis. This paper will give to the researchers an introduction to what they need to know and understand about the networks, WSNs, and IoT applications from scratch. Thus, the fundamental purpose of this paper is to introduce research trends and recent work on the use of IoT technology and the conclusion that has been reached as a result of undertaking the Ph.D. study.

show abstract

Section: Aspects Of Energy-efficiency Optimization Methodsmentioning

confidence: 99%

Energy Efficiency for Green Internet of Things (IoT) Networks: A Survey

Farhan

Hameed

Ahmed

et al. 2021

Network

View full text Add to dashboard Cite

show abstract

“…A number of use cases have been considered to capture different smart environments in the IoT, however, they do not address the impact of the networking power consumption which occurs to establish links between virtual functions i.e., VMs. In another study, the authors of [27], tackle the energy efficiency of machine learning tools by optimizing the software code for devices with limited computation and energy so that it can be deployed on the IoT and edge. In a comprehensive study, the authors of [43] provide a detailed survey and review about the recent advances towards the goal of enabling efficient DNN processing.…”

Section: Related Workmentioning

confidence: 99%

“…It is reported that billions of smart objects ranging from RFID tags to smart TVs and vehicles have already been connected to the Internet and their numbers are continually on the rise [47]. IoT alone is expected to produce 5 quintillions of data on a daily basis and driverless cars are reported to generate 4TB of data during a single hour of driving per day [27]. This abundance of sensory data makes AI and DNN models particularly attractive for increased deployments in many areas.…”

Section: Introductionmentioning

confidence: 99%

Energy-Efficient AI over a Virtualized Cloud Fog Network

Yosuf

Mohamed

Alenazi

et al. 2021

Proceedings of the Twelfth ACM International Conference on Future Energy Systems

View full text Add to dashboard Cite

Deep Neural Networks (DNNs) have served as a catalyst in introducing a plethora of next-generation services in the era of Internet of Things (IoT), thanks to the availability of massive amounts of data collected by the objects on the edge. Currently, DNN models are used to deliver many Artificial Intelligence (AI) services that include image and natural language processing, speech recognition, and robotics. Accordingly, such services utilize various DNN models that make it computationally intensive for deployment on the edge devices alone. Thus, most AI models are offloaded to distant cloud data centers (CDCs), which tend to consolidate large amounts of computing and storage resources into one or more CDCs. Deploying services in the CDC will inevitably lead to excessive latencies and overall increase in power consumption. Instead, fog computing allows for cloud services to be extended to the edge of the network, which allows for data processing to be performed closer to the end-user device. However, different from cloud data centers, fog nodes have limited computational power and are highly distributed in the network. In this paper, using Mixed Integer Linear Programming (MILP), we formulate the placement of DNN inference models, which is abstracted as a network embedding problem in a Cloud Fog Network (CFN) architecture, where power savings are introduced through trade-offs between processing and networking. We study the performance of the CFN architecture by comparing the energy savings when compared to the baseline approach which is the CDC. CCS CONCEPTS• Computing methodologies → Machine learning; • Networks → Network architectures.

show abstract

“…Machine learning at the edge has recently attracted significant attention, both in industry-based and academic research [3], [4]. A number of ML hardware accelerator platforms, such as Google's Edge TPU, have recently been developed, and a number of papers have evaluated them.…”

Section: Related Workmentioning

confidence: 99%

Exploring Deep Neural Networks on Edge TPU

Hosseininoorbin¹,

Layeghy²,

Kusý³

et al. 2021

Preprint

View full text Add to dashboard Cite

This paper explores the performance of Google's Edge TPU on feed forward neural networks. We consider Edge TPU as a hardware platform and explore different architectures of deep neural network classifiers, which traditionally has been a challenge to run on resource constrained edge devices. Based on the use of a joint-time-frequency data representation, also known as spectrogram, we explore the trade-off between classification performance and the energy consumed for inference. The energy efficiency of Edge TPU is compared with that of widely-used embedded CPU ARM Cortex-A53. Our results quantify the impact of neural network architectural specifications on the Edge TPU's performance, guiding decisions on the TPU's optimal operating point, where it can provide high classification accuracy with minimal energy consumption. Also, our evaluations highlight the crossover in performance between the Edge TPU and Cortex-A53, depending on the neural network specifications. Based on our analysis, we provide a decision chart to guide decisions on platform selection based on the model parameters and context.

show abstract

Energy-Efficient Machine Learning on the Edges

Cited by 23 publications

References 36 publications

Energy Efficiency for Green Internet of Things (IoT) Networks: A Survey

Energy Efficiency for Green Internet of Things (IoT) Networks: A Survey

Energy-Efficient AI over a Virtualized Cloud Fog Network

Exploring Deep Neural Networks on Edge TPU

Contact Info

Product

Resources

About