An energy efficient FPGA partial reconfiguration based micro-architectural technique for IoT applications

Kiat, Wei Pau; Mok, Kai-Ming; Lee, Wai-Kong; Goh, Hock Guan; Achar, R.

doi:10.1016/j.micpro.2019.102966

Cited by 29 publications

(9 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The use of FPGA based CNN accelerator is becoming a trend recently. This is also in line with the emerging trend in using FPGA based processor [28] for IoT applications. The proposed DoubleQExt technique is useful for IoT applications, in which the memory size and accuracy can be tradeoff according to the requirements (i.e., high accuracy or low memory consumption).…”

Section: Comparison With Related Worksupporting

confidence: 82%

See 1 more Smart Citation

DoubleQExt: Hardware and Memory Efficient CNN Through Two Levels of Quantization

See

Ng²,

Tan

et al. 2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

To fulfil the tight area and memory constraints in IoT applications, the design of efficient Convolutional Neural Network (CNN) hardware becomes crucial. Quantization of CNN is one of the promising approach that allows the compression of large CNN into a much smaller one, which is very suitable for IoT applications. Among various proposed quantization schemes, Power-of-two (PoT) quantization enables efficient hardware implementation and small memory consumption for CNN accelerators, but requires retraining of CNN to retain its accuracy. This paper proposes a two-level post-training static quantization technique (DoubleQ) that combines the 8-bit and PoT weight quantization. The CNN weight is first quantized to 8-bit (level one), then further quantized to PoT (level two). This allows multiplication to be carried out using shifters, by expressing the weights in their PoT exponent form. DoubleQ also reduces the memory storage requirement for CNN, as only the exponent of the weights is needed for storage. However, DoubleQ trades the accuracy of the network for reduced memory storage. To recover the accuracy, a selection process (DoubleQExt) was proposed to strategically select some of the less informative layers in the network to be quantized with PoT at the second level. On ResNet-20, the proposed DoubleQ can reduce the memory consumption by 37.50% with 7.28% accuracy degradation compared to 8-bit quantization. By applying DoubleQExt, the accuracy is only degraded by 1.19% compared to 8-bit version while achieving a memory reduction of 23.05%. This result is also 1% more accurate than the state-of-the-art work (SegLog). The proposed DoubleQExt also allows flexible configuration to trade off the memory consumption with better accuracy, which is not found in the other state-of-the-art works. With the proposed two-level weight quantization, one can achieve a more efficient hardware architecture for CNN with minimal impact to the accuracy, which is crucial for IoT applications.

show abstract

Section: Comparison With Related Worksupporting

confidence: 82%

“…On the other hand, FPGA is also widely used to implement IoT sensor nodes, due to its field reconfigurability [28]. In this case, the deep learning network size may not be the most critical issue, since we can load the network into a large DDR RAM.…”

Section: Applying Doubleqext To Iot Sensor Nodesmentioning

confidence: 99%

DoubleQExt: Hardware and Memory Efficient CNN Through Two Levels of Quantization

See

Ng²,

Tan

et al. 2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…Kiat et al presents a reconfigurable reduced instruction set computing (RISC) processor to meet design goals of IoT applications. The FPGA based processor can be reconfigured into either multi-cycle execution mode or pipeline execution mode depending on dynamic workload requirement of IoT applications [173].…”

Section: Wireless Sensor Network (Wsn)mentioning

confidence: 99%

Green Internet of Things (GIoT): Applications, Practices, Awareness, and Challenges

et al. 2021

View full text Add to dashboard Cite

Internet of things (IoT) is one of key pillars in fifth generation (5G) and beyond 5G (B5G) networks. It is estimated to have 42 billion IoT devices by the year 2025. Currently, carbon emissions and electronic waste (e-waste) are significant challenges in the information & communication technologies (ICT) sector. The aim of this article is to provide insights on green IoT (GIoT) applications, practices, awareness, and challenges to a generalist of wireless communications. We garner various efficient enablers, architectures, environmental impacts, technologies, energy models, and strategies, so that a reader can find a wider range of GIoT knowledge. In this article, various energy efficient hardware design principles, data-centers, and software based data traffic management techniques are discussed as enablers of GIoTs. Energy models of IoT devices are presented in terms of data communication, actuation process, static power dissipation and generated power by harvesting techniques for optimal power budgeting. In addition, this article presents various effective behavioral change models and strategies to create awareness about energy conservation among users and service providers of IoTs. Fog/Edge computing offers a platform that extends cloud services at the edge of network and hence reduces latency, alleviates power consumption, offers improved mobility, bandwidth, data privacy, and security. Therefore, we present the energy consumption model of a fog-based service under various scenarios. Problems related to ever increasing data in IoT networks can be solved by integrating artificial intelligence (AI) along with machine learning (ML) models in IoT networks. Therefore, this article provides insights on role of the ML in the GIoT. We also present how legislative policies support adoption of recycling process by various stakeholders. In addition, this article is presenting future research goals towards energy efficient hardware design principles and a need of coordination between policy makers, IoT devices manufacturers along with service providers.INDEX TERMS Fifth generation, Internet of Things, green Internet of Things, fog, machine learning.

show abstract

“…The simulation findings show that the suggested system has dramatically increased connectivity efficiency in network output by an average of 13.5%, a decrease of 38.5% in packets, a drop of 13.5% for network latency, 16% for energy usage, and 26% for smart agriculture routing overhead in contrast with other strategies. Kiat, Wei-Pau Mok, et al [77] proposed implementing an energy-efficient technique used in IoT applications to support varying performance and energy trade-offs utilizing a reduced instruction set computing (RISC) processor. The processor can be reconfigured into pipeline mode and multi-cycle mode (high computer speed, at the cost of high dynamic power).…”

Section: Energy Management Surveymentioning

confidence: 99%

Energy Management for Internet of Things via Distributed Systems

Sadeeq

Zeebaree

2021

JASTT

103

View full text Add to dashboard Cite

The distributed energy system (DES) architecture is subject to confusion about renewable energy limits, primary energy supply and energy carriers' costs. For the grid to use unreliable electricity sources, the end-user's on-demand presence in the intelligent energy management context is essential. The participation of end-users could influence the management of the system and the volatility of energy prices. By delivering auxiliary services using demand side-resource to increase system reliability, robust planning, constraint control and scheduling, consumers may support grid operators. The optimized approach to managing energy resources enhances demand response to renewable energy sources integrally, controls the demand curve with load versatility as the system requires it. The opportunity to adjust/regulate the charging profile by choosing a particular device. This article discusses a literature and policy analysis that looks at the role of energy management system aggregators and the end-users participating in subsidiary systems within Smart Grid programmers and technologies. In the implementation of aggregators for energy management systems, the objective is to understand the patterns, threats, obstacles and potential obstacles.

show abstract

An energy efficient FPGA partial reconfiguration based micro-architectural technique for IoT applications

Cited by 29 publications

References 15 publications

DoubleQExt: Hardware and Memory Efficient CNN Through Two Levels of Quantization

DoubleQExt: Hardware and Memory Efficient CNN Through Two Levels of Quantization

Green Internet of Things (GIoT): Applications, Practices, Awareness, and Challenges

Energy Management for Internet of Things via Distributed Systems

Contact Info

Product

Resources

About