检索结果-内蒙古大学图书馆

IEEE International Conference on Communications (IEEE ICC)

作者： Tamburello, Marialaura Caruso, Giuseppe Adami, Davide Giordano, Stefano Univ Pisa Dept Informat Engn Pisa Italy Univ Pisa CNIT Pisa Italy

ISBN: (纸本)9781538674628

Recent impressive growth of AI applications in the most diversified heterogeneous domains is largely motivated by the availability of hardware accelerators used from the backstage of data centers (such as TPU, Tensor processing Units, or VPUs, Visual processing Units) to the far edge of embedded devices equipped with DPUs and Deep Learning processing Units. High level toolchains for a more friendly usability of these platform had similar relevance in the process. In this paper we considered edge devices that provide an essential contribution for the deployment of "distributed intelligence" and are used typically at the gateway, CPE or Edge computing level. One of the typical assumptions is that Field Programmable Gate Array (FPGA) are far more expensive - with respect to power consumption - than legacy SBCs (single board computers). The main contribution of the paper is a fair comparison (at the same clock frequency and with the same main CPU) of processing time and power consumption of two different boards used for deep neural network classification. We will highlight the relevance of classification speed with respect to common KPIs adopted to compare the performances of automatic classification such as Loss, Precision, Recall, etc. This will be particularly relevant in the challenging domains of hardware accelerated real time control loops to provide distributed intelligence at the application level but also at the inner functions of emerging networking architectures.

关键词： Artificial Intelligence (AI) classification Convolutional neural network (CNN) power consumption Deep Learning processing Unit (DPU)

来源：评论

学校读者我要写书评

暂无评论

To Think Like a Vertex (or Not) for distributed Training of Graph neural networks 23

To Think Like a Vertex (or Not) for Distributed Training of ...

引用

23rd IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGrid)

作者： Kulkarni, Varad Chaturvedi, Akarsh Naman, Pranjal Simmhan, Yogesh Indian Inst Sci Bangalore 560012 India

ISBN: (纸本)9798350302080

Graph neural networks (GNNs) train neural networks that combine the topological properties of a graph with the vertex and edge features to perform tasks such as node classification and link prediction. We propose a novel middleware that approaches GNN training from the perspective of a vertex-centric model (VCM) of distributed graph processing and overlays neural network training over it. Giraph Graph neural network (G2N2) uses a three-phase execution pattern by construction a distributed computation graph per mini-batch, and maps the forward and backward passes of the GNN training to VCM. We implement a prototype of G2N2 in Apache Giraph and report results from a preliminary evaluation using two real-world graphs on a commodity cluster.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Control Strategy for DC Soft Open Points in Large-scale Distribution networks with distributed Generators

引用

CSEE Journal of Power and Energy Systems 2022年第3期8卷 732-742页

作者： Dejan Ivic Predrag Stefanov School of Electrical Engineering University of Belgrade11120 BelgradeSerbia

Development of power electronic devices and their implementation in power distribution networks significantly improves not only the easier integration of distributed generation into distribution networks but also the further refinement of control algorithms specialized in the application of power management procedures in distribution *** of DC Soft Open Points(DCSOP)for closing loops in distribution networks could serve as one of the most representative examples of the use of power electronic devices as management resources in modern distribution *** paper describes the possibilities of the application of DCSOP used for loops closure and power flow control in medium voltage distribution networks with integrated distributed *** all the benefits of loops closure and the controllability of the DCSOP,it is considered as a management resource capable of achieving the desired/optimal state in the distribution *** management strategy proposed in this paper covers two cases:the first case is where an adequate communication infrastructure is available when all DCSOP operations are based on Optimal Power Flow(OPF)calculations,and the second case is when OPF results are unavailable for some reason(most often due to communication interruptions)and DCSOP operations are based on processing available data using a neural network-based *** proposed management strategy has been implemented and tested on an IEEE 33 distribution test network with integrated distributed *** simulation results,described in this paper,can serve not only as a basis for further scientific research but also as a basis for the practical implementation of real controllers based on the proposed algorithm and its application in practice,and in real distribution networks.

关键词： DC soft open points(DCSOP) neural networks optimal power flow(OPF) optimization in distribution networks

来源：评论

学校读者我要写书评

暂无评论

Synapse Compression for Event-Based Convolutional-neural-network Accelerators

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2023年第4期34卷 1227-1240页

作者： Bamberg, Lennart Pourtaherian, Arash Waeijen, Luc Chahar, Anupam Moreira, Orlando GrAI Matter Labs F-75012 Paris France

Manufacturing-viable neuromorphic chips require novel compute architectures to achieve the massively parallel and efficient information processing the brain supports so effortlessly. The most promising architectures for that are spiking/event-based, which enables massive parallelism at low complexity. However, the large memory requirements for synaptic connectivity are a showstopper for the execution of modern convolutional neural networks (CNNs) on massively parallel, event-based architectures. The present work overcomes this roadblock by contributing a lightweight hardware scheme to compress the synaptic memory requirements by several thousand times-enabling the execution of complex CNNs on a single chip of small form factor. A silicon implementation in a 12-nm technology shows that the technique achieves a total memory-footprint reduction of up to 374x compared to the best previously published technique at a negligible area overhead.

关键词： CNN compression dataflow event-based hardware accelerator near-memory compute neuromorphic sparsity spiking

来源：评论

学校读者我要写书评

暂无评论

Dynamic Event-Triggered Approach for distributed State and Parameter Estimation Over networks Subjected to Deception Attacks

引用

IEEE TRANSACTIONS ON SIGNAL AND INFORMATION processing OVER networkS 2023年 9卷 373-385页

作者： Basit, Abdul Tufail, Muhammad Rehan, Muhammad Ahn, Choon Ki Pakistan Inst Engn & Appl Sci PIEAS Dept Elect Engn Islamabad 45650 Pakistan Korea Univ Sch Elect Engn Seoul 136701 South Korea

This study is focused on addressing the dynamic event-triggered distributed state and unknown parameter estimation problem for discrete-time nonlinear systems that have known linear dynamics and unknown nonlinearities and are subject to deception attacks. A neural-network-based unified estimation framework is introduced to estimate the unknown nonlinear function in conjunction with the system state and unknown parameters. Each sensor uses its own measurements and data from the neighboring sensors to calculate the overall estimates. The information-sharing network is assumed to be vulnerable to deception attacks, which are modeled using a Bernoulli distributed random variable. Additionally, a dynamic event-triggered strategy is adopted to alleviate resource consumption. Based on Lyapunov theory, the stability of the unified estimation framework is proven in terms of the uniformly ultimately bounded error. Moreover, the design conditions for the estimator are presented in the form of matrix inequalities. Finally, a simulation example is presented to demonstrate the effectiveness of the proposed framework.

关键词： Wireless sensor networks event-triggered deception attacks state estimation unknown parameter

来源：评论

学校读者我要写书评

暂无评论

Non-Linear Coded Computation for distributed CNN Inference: A Learning-based Approach 44

Non-Linear Coded Computation for Distributed CNN Inference: ...

引用

44th IEEE International Conference on distributed Computing Systems (ICDCS)

作者： Wang, Haoyu Liu, Yiyu Xu, Chunhui Liu, Xing Tan, Ming Southern Univ Sci & Technol Dept Comp Sci & Engn Shenzhen Peoples R China

ISBN: (纸本)9798350354720;9798350354713

distributed inference enables multiple distributed computing devices to cooperatively perform statistical inference or data analysis tasks in order to reduce the inference delay. Due to the device heterogeneity in terms of processing capacities, some devices may experience slower processing speeds or even fail to compute the inference task. This leads to a larger inference delay or sometimes inference task failure. In this work, we propose a novel learning-based coded inference framework for convolutional neural networks (CNNs) by introducing redundant computing devices. By using an encoder to determine the inference inputs of redundant devices and an decoder to reconstruct the convolutional segment result of the CNNs, we show that our framework can significantly enhance the resilience and efficiency of distributed inference systems. By deploying a robust encoding-decoding scheme, our framework dynamically compensates for devices with sub-optimal performance or computational failures. This ensures the overall system to maintain high levels of accuracy and reduce inference latency, even in heterogeneous device environments. Our evaluation results reveal the profound impact of our proposed framework on enhancing distributed inference tasks for CNNs. Based on the experiments conducted on image classification across various realistic datasets, we have demonstrated that our framework maintains high accuracy levels, even in scenarios where some distributed devices encounter failures, validating the robustness of our proposed approach.

关键词： distributed inference coded computation convolutional neural network fault tolerance redundancy

来源：评论

学校读者我要写书评

暂无评论

Next-generation IIoT security: Comprehensive comparative analysis of CNN-based approaches

引用

KNOWLEDGE-BASED SYSTEMS 2025年 316卷

作者： Dong, Huiyao Kotenko, Igor Levshun, Dmitry ITMO Univ Kronverksky Pr 49Bldg A St Petersburg 197101 Russia Russian Acad Sci SPC RAS St Petersburg Fed Res Ctr 3914th Liniya St Petersburg 199178 Russia

Industrial Internet of Things (IIoT) presents a range of benefits but also introduces security vulnerabilities. This paper systematically compares different deep learning model structures for IIoT intrusion detection. Four Convolutional neural network (CNN) classifiers are implemented, including hybrid CNN-GRU, 1D Xception, and 1D Resnet models. The evaluation focuses on robustness and generalizability across two datasets to assess overall performance and detection rates. To handle imbalanced data, preprocessing involves PCA feature selection and hybrid resampling techniques. Furthermore, we design a distributed training process for massive datasets and continuous learning, enabling efficient large-scale processing. For the experimental evaluation, we make use of two datasets with multi-label classification tasks and analyze performance metrics of different models in terms of their detection abilities and efficiency. The proposed methods demonstrate strong performance, successfully identifying even rare attacks. In addition, we conduct a performance comparison with existing publications that utilize the same datasets, confirming that our models are on par with state-of-the-art IIoT models.

关键词： Information security Artificial intelligence Deep learning State assessment Attack detection network traffic Industrial ioT

来源：评论

学校读者我要写书评

暂无评论

Microwave Photonics-Based Small Hydrostatic Pressure Sensing Assisted by Convolutional neural network

引用

IEEE Sensors Journal 2025年第11期25卷 20948-20955页

作者： Li, Songlin Xue, Ting Wu, Yan Li, Zhuping Wu, Bin Tianjin University Tianjin Key Laboratory of Process Measurement and Control School of Electrical and Information Engineering Tianjin300072 China Tianjin University State Key Laboratory of Precision Measuring Technology and Instruments Tianjin300072 China

Hydrostatic pressure measurement is essential in many industries, such as oil and gas production, chemical processing, and environmental monitoring. Due to the minimal impact of small hydrostatic pressure on ordinary silica optical fibers, research on its measurement utilizing optical fiber sensing technology remains limited. In this paper, a novel microwave photonics technique, termed optical carrier-based microwave interferometry (OCMI), is utilized for small hydrostatic pressure sensing with the assistance of a convolutional neural network (CNN). The theory of OCMI-based phase demodulation is established, and numerical simulations are conducted to investigate the factors affecting the axial displacement of the fiber core. In practical experiments, the phase demodulation method is applied to small hydrostatic pressure measurements;however, the results are suboptimal. Therefore, the CNN is developed to assist in the implementation of accurate small hydrostatic pressure sensing. The small hydrostatic pressures predicted by the well-trained CNN model are in good agreement with the actual values, with an error of less than 0.25 kPa. In addition, the prediction results from multiple Fabry-Perot interferometers demonstrate the feasibility and effectiveness of utilizing CNN for OCMI-based small hydrostatic pressure sensing. The introduction of machine learning broadens the application scope of the OCMI technique, allowing it to be employed for distributed sensing of a wider range of physical, chemical, and biological quantities. © 2001-2012 IEEE.

关键词： Interferometry

来源：评论

学校读者我要写书评

暂无评论

The Capability of Wavelet Convolutional neural network for Detecting Cyber Attack of distributed Denial of Service in Smart Grid 18

The Capability of Wavelet Convolutional Neural Network for D...

引用

18th International Computer Conference on Wavelet Active Media Technology and Information processing, ICCWAMTIP 2021

作者： Monday, Happy Nkanta Li, Jian Ping Nneji, Grace Ugochi Yutra, Abel Zenebe Lemessa, Bona Debela Nahar, Saifun James, Edidiong Christopher Haq, Amin Ul University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu Sichuan 611731 China University of Electronic Science and Technology of China School of Information and Software Engineering Chengdu Sichuan 611731 China University of Missouri School of Computer and Information Systems St LouisMO United States

ISBN: (纸本)9781665413640

The electrical system's dependability, security, and efficiency are all improved through smart grid technologies. Its dependence on digital communication technology, on the other hand, introduces new risks and vulnerabilities that should be examined for the purpose to providing effective and trustworthy service delivery. This study presents a method for the detection of distributed denial of service (DDoS) attacks on smart grid infrastructure. Continuous wavelet transform (CWT) is used in the suggested approach to convert one-dimensional traffic data to two-dimensional time-frequency domain scalogram as the input to the wavelet convolutional neural network (WavCovNet) to detect anomalous behavior in the data by distinguishing attack features from normal patterns. Our results demonstrate that the proposed approach detects DDoS attacks with a high rate of detection and with a very low rate of false alarm. © 2021 IEEE.

关键词： Smart power grids

来源：评论

学校读者我要写书评

暂无评论

Enhancing μNAS for 1D CNNs on Microcontrollers 55th

Enhancing μNAS for 1D CNNs on Microcontrollers

引用

55th Meeting of the Italian Electronics Society

作者： Capello, Alessio Berta, Riccardo Ballout, Hadi Fresta, Matteo Soltanmuradov, Vafali Bellotti, Francesco Univ Genoa Dept Elect Elect & Telecommun Engn DITEN Via Opera Pia 11a I-16145 Genoa Italy

ISBN: (纸本)9783031715204;9783031715181;9783031715174

Deep Learning (DL) has proved effective in a variety of application domains. However, the computational and memory demand posed by deep neural models limits the spread of DL on resource-constrained devices such as microcontrollers. An opportunity to tailor DL models to specific hardware constraints is given by neural Architecture Search (NAS), which automatically traverses a large search space, seeking for optimal architectures both in terms of hardware and performance, based on user specifications. State of the art open-source NAS tools for microcontrollers only support 2D Convolutional neural network (CNN) and Multi Layer Perceptron (MLP), but do not consider 1D convolution, which is key for time series analysis and signal processing. This study focuses on enhancing the state-of-the-art mu NAS framework, by adding support for 1D CNN. Preliminary tests on a dummy dataset consisting of simple gaussian-distributed waveforms, demonstrate the system ability to find appropriate architectures to satisfy the specified constraints.

关键词： neural Architecture Search (NAS) Microcontrollers Optimization 1D Convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：