Double-Shift: A Low-Power DNN Weights Storage and Access Framework based on Approximate Decomposition and Quantization

2022 ◽  
Vol 27 (2) ◽  
pp. 1-16
Author(s):  
Ming Han ◽  
Ye Wang ◽  
Jian Dong ◽  
Gang Qu

One major challenge in deploying Deep Neural Network (DNN) in resource-constrained applications, such as edge nodes, mobile embedded systems, and IoT devices, is its high energy cost. The emerging approximate computing methodology can effectively reduce the energy consumption during the computing process in DNN. However, a recent study shows that the weight storage and access operations can dominate DNN's energy consumption due to the fact that the huge size of DNN weights must be stored in the high-energy-cost DRAM. In this paper, we propose Double-Shift, a low-power DNN weight storage and access framework, to solve this problem. Enabled by approximate decomposition and quantization, Double-Shift can reduce the data size of the weights effectively. By designing a novel weight storage allocation strategy, Double-Shift can boost the energy efficiency by trading the energy consuming weight storage and access operations for low-energy-cost computations. Our experimental results show that Double-Shift can reduce DNN weights to 3.96%–6.38% of the original size and achieve an energy saving of 86.47%–93.62%, while introducing a DNN classification error within 2%.

2020 ◽  
pp. 1042-1057
Author(s):  
Xiaojing Hou ◽  
Guozeng Zhao

With the wide application of the cloud computing, the contradiction between high energy cost and low efficiency becomes increasingly prominent. In this article, to solve the problem of energy consumption, a resource scheduling and load balancing fusion algorithm with deep learning strategy is presented. Compared with the corresponding evolutionary algorithms, the proposed algorithm can enhance the diversity of the population, avoid the prematurity to some extent, and have a faster convergence speed. The experimental results show that the proposed algorithm has the most optimal ability of reducing energy consumption of data centers.


Internet of Things (IoT) constitutes a network of various devices has an equipment with the mandatory facility of communication and optional facilities of sensing, information collecting, storage and processing. IoT network has been used for research and development purpose in many application areas such as military environment, traffic management, and e-healthcare system. IoT network was enormous in scale and complexity, mainly in terms of energy efficiency because battery lifetime is limited. The previous routing protocols for IoT are difficult and require a huge memory use and high energy consumption which are insufficient for IoT network processing. For that reason, an efficient routing algorithm needed to decrease energy consumption while communication. To tackle this problem, this paper proposes Less Energy Consumption Routing (LECR) algorithm. This algorithm reduces energy consumption using 4 ways in IoT, (1) Sleep and Wake up Scheduling, (2) Route Discovery in IoT Base Station (3) Less Power Consumption Route for Communication (4) Reduce Overhead while Routing. The experimental result proves the LECR algorithm reduces IoT devices battery drain and increases lifetime of the IoT network efficiently


Sensors ◽  
2019 ◽  
Vol 19 (23) ◽  
pp. 5106 ◽  
Author(s):  
Taewon Song ◽  
Taeyoon Kim

Internet of Things (IoT) technology is rapidly expanding the use of its application, from individuals to industries. Owing to this, the number of IoT devices has been exponentially increasing. Considering the massive number of the devices, overall energy consumption is becoming more serious. From this point of view, attaching low-power wake-up radio (WUR) to the devices can be one of the candidate solutions to deal with this problem. With WUR, IoT devices can go to sleep until WUR receives a wake-up signal, which enables a significant reduction of its power consumption. Meanwhile, one concern for WUR operation is the addressing mechanism, since operational efficiency of the wake-up feature can significantly vary depending on the addressing mechanism. We therefore introduce addressing mechanisms for IoT devices equipped with WUR and analyze their performances, such as elapsed time to wake up, false positive probability and power/energy consumption, to provide appropriate addressing mechanisms over practical environments for IoT devices with WUR.


2020 ◽  
Author(s):  
Marcelo Brandalero ◽  
Luigi Carro ◽  
Antonio Carlos Schneider Beck

With recent changes in transistor scaling trends, the design of all types of processing systems has become increasingly constrained by power consumption. At the same time, driven by the needs of fast response times, many applications are migrating from the cloud to the edge, pushing for the challenge of increasing the performance of these already power-constrained devices. The key to addressing this problem is to design application-specific processors that perfectly match the application's requirements and avoid unnecessary energy consumption. However, such dedicated platforms require significant design time and are thus unable to match the pace of fast-evolving applications that are deployed in the Internet-of-Things (IoT) every day. Motivated by the need for high energy efficiency and high flexibility in hardware platforms, this thesis paves the way to a new class of low-power adaptive processors that can achieve these goals by automatically modifying their structure at run time to match different applications' resource requirements. The proposed Multi-Target Adaptive Reconfigurable Architecture (MuTARe) is based upon a Coarse-Grained Reconfigurable Architecture (CGRA) that can transparently accelerate already-deployed applications, but incorporates novel compute paradigms such as Approximate Computing (AxC) and Near-Threshold Voltage Computing (NTC) to improve its efficiency. Compared to a traditional system of heterogeneous processing cores (similar to ARM's big.LITTLE), the base MuTARe architecture can (without any change to the existing software) improve the execution time by up to $1.3\times$, adapt to the same task deadline with $1.6\times$ smaller energy consumption or adapt to the same low energy budget with $2.3\times$ better performance. When extended for AxC, MuTARe's power savings can be further improved by up to $50\%$ in error-tolerant applications, and when extended for NTC, MuTARe can save further $30\%$ energy in memory-intensive workloads.


Electronics ◽  
2020 ◽  
Vol 9 (12) ◽  
pp. 2191
Author(s):  
Waqas Tariq Toor ◽  
Maira Alvi ◽  
Mamta Agiwal

This paper focuses on proposing a new access barring scheme for internet of things (IoT) devices in long term evolution advanced (LTE/LTE-A) and 5G networks. Massive number of IoT devices communicating simultaneously is one of the hallmarks of the future communication networks such as 5G and beyond. The problem of congestion also comes with this massive communication for which access barring is one of the solutions. So, it is required that sophisticated access barring techniques are designed such that the congestion is avoided and these devices get served in less time. Legacy access barring schemes like access class barring (ACB) and extended access barring (EAB) suffer from high energy consumption and high access delay respectively. However, our proposed scheme provides less energy consumption than ACB while giving less access delay than EAB. The proposed scheme maximizes the success probability while reducing the number of collisions at the same time. The scheme is based on an approximation of the number of IoT devices based on details available to the eNodeB of the number of idle, successful and collided preambles. Extensive Matlab simulations are performed to validate our claims and analysis.


2021 ◽  
Vol 17 (3) ◽  
pp. 1-28
Author(s):  
Yunji Liang ◽  
Xin Wang ◽  
Zhiwen Yu ◽  
Bin Guo ◽  
Xiaolong Zheng ◽  
...  

With the proliferation of Internet of Things (IoT) devices in the consumer market, the unprecedented sensing capability of IoT devices makes it possible to develop advanced sensing and complex inference tasks by leveraging heterogeneous sensors embedded in IoT devices. However, the limited power supply and the restricted computation capability make it challenging to conduct seamless sensing and continuous inference tasks on resource-constrained devices. How to conduct energy-efficient sensing and perform rich-sensor inference tasks on IoT devices is crucial for the success of IoT applications. Therefore, we propose a novel energy-efficient collaborative sensing framework to optimize the energy consumption of IoT devices. Specifically, we explore the latent correlations among heterogeneous sensors via an attention mechanism in temporal convolutional network to quantify the dependency among sensors, and characterize the heterogeneous sensors in terms of energy consumption to categorize them into low-power sensors and energy-intensive sensors . Finally, to decrease the sampling frequency of energy-intensive sensors , we propose a multi-task learning strategy to predict the statuses of energy-intensive sensors based on the low-power sensors . To evaluate the performance of the proposed collaborative sensing framework, we develop a mobile application to collect concurrent heterogeneous data streams from all sensors embedded in Huawei Mate 8. The experimental results show that latent correlation learning is greatly helpful to understand the latent correlations among heterogeneous streams, and it is feasible to predict the statuses of energy-intensive sensors by low-power sensors with high accuracy and fast convergence. In terms of energy consumption, the proposed collaborative sensing framework is able to preserve the energy consumption of IoT devices by nearly 50% for continuous data acquisition tasks.


Water ◽  
2020 ◽  
Vol 12 (4) ◽  
pp. 1204 ◽  
Author(s):  
Alexandra Siatou ◽  
Anthoula Manali ◽  
Petros Gikas

The high-energy consumption of wastewater treatment plants (WWTPs) is a crucial issue for municipalities worldwide. Most WWTPs in Greece operate as extended aeration plants, which results in high operational costs due to high energy needs. The present study investigated the energy requirements of 17 activated sludge WWTPs in Greece, serving between 1100–56,000 inhabitants (population equivalent, PE), with average daily incoming flowrates between 300–27,300 m3/d. The daily wastewater production per inhabitant was found to lie between 0.052 m3/PE·d and 0.426 m3/PE·d, with average volume of 0.217 ± 0.114 m3/PE·d. The electric energy consumption per volume unit (EQ (kWh/m3)) was between 0.128–2.280 kWh/m3 (average 0.903 ± 0.509 kWh/m3) following a near logarithmic descending correlation with the average incoming flowrate (Qav) (EQ = −0.294lnQav + 3.1891; R2 = 0.5337). A similar relationship was found between the daily electric energy requirements for wastewater treatment per inhabitant (EPE (kWh/PE·d)) as a function of PE, which varied from 0.041–0.407 kWh/PE·d (average 0.167 ± 0.101 kWh/PE·d)) (EPE = −0.073ln(PE) + 0.8425; R2 = 0.6989). Similarly, the daily energy cost per inhabitant (E€/PE (€/PE·d)) as a function of PE and the electric energy cost per wastewater volume unit (E€/V (€/m3)) as a function of average daily flow (Qav) were found to follow near logarithmic trends (E€/PE = −0.013ln(PE) + 0.1473; R2 = 0.6388, and E€/V = −0.052lnQav + 0.5151; R2 = 0.6359), respectively), with E€/PE varying between 0.005–0.073 €/PE·d (average 0.024 ± 0.019 €/PE·d) and E€/V between 0.012–0.291 €/m3 (average 0.111 ± 0.077 €/m3). Finally, it was calculated that, in an average WWTP, the aeration process is the main energy sink, consuming about 67.2% of the total electric energy supply to the plant. The large variation of energy requirements per inlet volume unit and per inhabitant served, indicate that there is large ground for improving the performance of the WWTPs, with respect to energy consumption.


2012 ◽  
Vol 511 ◽  
pp. 64-69
Author(s):  
Pei Zhang ◽  
Han Zhu ◽  
Apostolos Fafitis

Energy consumption and CO2 emissions in buildings is becoming an increasingly important issue. Steel is a major building material with high energy cost. In a reinforced concrete (RC) structure, it accounts for the maximum energy consumption. There is a need to quantify the steel amount in RC for various situations so that reduction or optimization in steel usage can be analyzed. In this paper two different calculations (Calculation-I and Calculation-II) are conducted by using two groups of steel in designing beams, columns and plates for a 20000 m2 five-storeyed frame RC structure. In Calculation-I, or Cal-I in abbreviation, the steel used for beams, columns and plates is HRB335, HRB400 and HPB235 respectively. In Calculation-II, or Cal-II in abbreviation, the steel used for beams, columns and plates is HRB400, HRB500 and CRB550 respectively. The strength of steel used in Cal-II is higher than that in Cal-I. The calculation is carried out by following the standardized concrete structural design code, and the steps involved in calculation are given in certain details as seen necessary. The corresponding energy for producing the steel used in beams, columns and plates is also computed and normalized on per square meter basis. The results show that Cal-II saves 101.76 tons of steel than Cal-I, or 5.09kg/m2, which means a saving of about 64.11 t of standard coal or 1.6×102 t CO2 for the whole structure, or 3.2 kg of standard coal or 7.98kg CO2 for per square meter.


2021 ◽  
Vol 2021 ◽  
pp. 1-14
Author(s):  
Felicia Engmann ◽  
Kofi Sarpong Adu-Manu ◽  
Jamal-Deen Abdulai ◽  
Ferdinand Apietu Katsriku

In Wireless Sensor Networks, sensor nodes are deployed to ensure continuous monitoring of the environment which requires high energy utilization during the data transmission. To address the challenge of high energy consumption through frequent independent data transmission, the IEEE 802.11b provides a backoff window that reduces collisions and energy losses. In the case of Internet of Things (IoTs), billions of devices communicate with each other simultaneously. Therefore, adapting the contention/backoff window size to data traffic to reduce congestion has been one such approach in WSN. In recent years, the IEEE 802.11b MAC protocol is used in most ubiquitous technology adopted for devices communicating in the IoT environment. In this paper, we perform a thorough evaluation of the IEEE 802.11b standard taking into consideration the channel characteristics for IoT devices. Our evaluation is aimed at determining the optimum parameters suitable for network optimization in IoT systems utilizing the IEEE 802.11b protocol. Performance analysis is made on the sensitivity of the IEEE 802.11b protocol with respect to the packet size, packet delivery ratio (PDR), end-to-end delay, and energy consumption. Our studies have shown that for optimal performance, IoT devices using IEEE 802.11b channel require data packet of size 64 bytes, a data rate of 11Mbps, and an interpacket generation interval of 4 seconds. The sensitivity analysis of the optimal parameters was simulated using NS3. We observed PDR values ranging between 27% and 31%, an average end-to-end delay ranging within 10-15 ms while the energy remaining was between 5.59 and 5.63Joules. The results clearly indicate that scheduling the rate of packet generation and transmission will improve the network performance for IoT devices while maintaining data reliability.


2020 ◽  
Author(s):  
Changyi Deng ◽  
Ruifeng Guo ◽  
Haotian Wu ◽  
Azhen Peng ◽  
Shaohua Du ◽  
...  

To solve the problem of high energy consumption and poor reliability of open CNC systems, through optimizing the slack time allocation issues, a set of real-time tasks in the open CNC system was researched to minimize energy consumption while maintaining the reliability of open CNC system. A Low Power and Reliability Based on Sliding Window (LPRSW) algorithm was proposed.


Sign in / Sign up

Export Citation Format

Share Document