检索结果-内蒙古大学图书馆

27th international European conference on parallel and distributed computing (Euro-Par)

作者： Choi, Kyusik Yang, Hoeseok Ajou Univ Dept ECE Suwon 16499 Gyeonggi South Korea

ISBN: (纸本)9783030856656;9783030856649

The model size and computation requirement of Deep Convolutional Neural networks (DNNs) have ever increased as their applications to various real-life use-cases, e.g., autonomous driving, are getting more pervasive and popular. While DNN workloads are executed on Graphics Processing Units (GPUs) in many cases, it is not trivial to improve the inference speed through the conventional DNN weight pruning techniques, due to the parallel architecture of GPUs. On the other hand, the coarse-grain pruning, also known as structured sparsity or structured pruning, can speedup the inference, but cause significant losses of accuracy. In this paper, we propose two fine-grain DNN pruning techniques that are aware of the underlying GPU architecture. For that, we analyze the hierarchical architecture of parallel processing elements and memory of GPU to identify the finest possible pruning where the removed weights can be safely skipped during the inference. The effectiveness of the proposed techniques has been evaluated with VGG16. Compared to the existing pruning techniques, the proposed methods result in significantly improved inference speed with less accuracy drop.

关键词： DNN Architecture-aware pruning GPU

来源：评论

学校读者我要写书评

暂无评论

Guest Editorial: Softwarized Networking for Next Generation Industrial Cyber-Physical Systems

引用

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS 2021年第8期17卷 5506-5510页

作者： Garg, Sahil Wang, Honggang Granelli, Fabrizio Li, Hongwei Univ Quebec Ecole Technol Super Montreal PQ H3C 1K3 Canada Univ Massachusetts Dartmouth N Dartmouth MA 02747 USA Univ Trento I-38122 Trento Italy Univ Elect Sci & Technol China Chengdu 610054 Peoples R China

The papers in this special section focus on softwarized networking for next generation industrial cyber-physical systems (CPSs). With the emergence of embedded and ubiquitous cyberphysical applications, the rationale of blending the physical and the virtual worlds has become ever promising. These papers examine several topics that are recently concerned in the community, including the software defined architectures and implementations, advanced machine learning and data analytics solutions, blockchain-based network services and applications, network function allocation, dependable and trustable solutions, energy efficient networks and services, and other enabling technologies for integrating softwarized networks into CPSs.

关键词： Special issues and sections Cyber-physical systems Informatics Data analysis Deep learning Resource management Industrial Internet of Things Computer architecture Blockchain Real-time systems Cloud computing

来源：评论

学校读者我要写书评

暂无评论

Deep learning methods for malicious URL detection using embedding techniques as Logistic Regression with Lasso penalty and Random Forest

Deep learning methods for malicious URL detection using embe...

引用

international conference on parallel distributed and Grid computing (PDGC)

作者： Isha Thakur Kajal Panda Sanjeev Kumar Cyber Security Technology Division Centre for Development of Advanced Computing Mohali India

ISBN: (纸本)9781665454025

In the Internet ecosystem, URLs (Uniform Resource Locators) are widely used to propagate malicious infections through spam, spear-phishing, drive-by-download exploitation, malware embedding, etc. The blacklisting, pattern analysis, and signature-matching approaches are widely used to detect such threats. However, these techniques are very effective in detecting known types of malicious URLs but unable to detect new types of attacks launched by the malicious URLs. Moreover, traditional machine learning-based malicious URL detection is highly dependent upon manual feature engineering, which is a costly approach. , This research study proposes deep learning-based malicious URL detection with less dependent on manual feature engineering to overcome the above-mentioned challenges. The Natural Language Processing (NLP) technique, i.e., Term Frequency-Inverse Document Frequency (TF-IDF) vectorizer, along with the N-gram parameter, is used for extracting the features. Afterward, two feature selection techniques as Embedded methods are used - Logistic Regression with L1 (Lasso) penalty and Random Forest Selection. Neural network models such as Deep Neural networks (DNN), Long Short-Term Memory (LSTM), and Convolution Neural networks (CNN) are constructed to detect malicious URLs. The experiment results demonstrate that the DNN model gives the best performance results as 96.95% accuracy, 99% precision, 100% recall, and 99% F1-score with logistic regression along with the L1 (Lasso) penalty as a feature selection method. Finally, this study compares state-of-art implemented deep learning methods for malicious URL detection.

关键词： Uniform resource locators Deep learning Biological system modeling Neural networks Manuals Feature extraction Natural language processing

来源：评论

学校读者我要写书评

暂无评论

Aspect Based Sentiment Classification Using Deep Memory Network

Aspect Based Sentiment Classification Using Deep Memory Netw...

引用

international conference on parallel distributed and Grid computing (PDGC)

作者： Jagendra Singh Penmetsa Dharmesh Adith Varma Sahas Gundapaneni Sai Preethi Bhupathiraju Darshiniram Mattaparthi Mohammad Sajid School of Computer Science Engineering and Technology Bennett University India Greater Noida India Department of Computer Science Aligarh Muslim University Aligarh India

ISBN: (纸本)9781665454025

Our main aim is to target and identify the polarity of the sentiments using a Deep memory network for the classification. And for emerging new technologies there is a drastic need for new services. Even though there are sequential neural methods like Long Short-Term Memory (LSTM) networks and a few other feature-based Support Vector Machine (SVM), they mostly approach focus on encapsulating the importance of every word in the sentence and at the same time interferes the aspect sentiment polarity. Hence with this growing number of services, the polarity is not enough to help people who seek these technologies. And using a deep memory network, provided new possibilities to go deeper into the research and helped to carry out Aspect based sentimental analysis. Through multiple computational layers, these degrees are equally given importance and have been calculated which makes it a neural attention model instead of maintaining focus on the external memory. This method proves that it is more sustainable than the other models, LSTM and SVM, and is also 15 times quicker with 9 computational layers.

关键词： Computational modeling Support vector machine classification Grid computing Long short term memory

来源：评论

学校读者我要写书评

暂无评论

Performance, Energy and parallelism: Using Near Data Processing in Utility and Cloud computing

Performance, Energy and Parallelism: Using Near Data Process...

引用

IEEE international conference on Utility and Cloud computing (UCC)

作者： Gul Agha Dipayan Mukherjee Atul Sandur Department of Computer Science University of Illinois at Urbana-Champaign Urbana IL USA

Massive amounts of data are generated by sensor networks, edge computers, IoT devices, and enterprise networks. To process this volume of data requires (1) a scalable programming model that is not only concurrent and distributed, but supports the mobility of data and processes (actors), and (2) algorithms to distribute computations between nodes in a manner that improves overall performance while considering energy use in the system. With appropriate programming tools, we can distribute a given computation in a way that makes effective use of edge devices to improve performance while lowering energy consumption. The paper describes our work building on ideas based on the Actor model of computation. These include characterizing the relation of performance and energy consumption in parallel computation, and methods to support scalable placement mechanisms under dynamically changing network conditions and computational loads on edge devices. The paper will conclude with a presentation with a summary of open research problems.

关键词： Performance evaluation Energy consumption Cloud computing Solid modeling Computational modeling distributed databases Programming

来源：评论

学校读者我要写书评

暂无评论

Trusted Auditing of Data Operation Behaviors in Cloud based on Blockchain and TEE

Trusted Auditing of Data Operation Behaviors in Cloud based ...

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Shujiang Xu Fansheng Wang Lianhai Wang Miodrag J. Mihaljević Shuhui Zhang Wei Shao Qihao Huang Key Laboratory of Computing Power Network and Information Security Ministry of Education Shandong Computer Science Center Qilu University of Technology (Shandong Academy of Sciences) Jinan China Mathematical Institute the Serbian Academy of Sciences and Arts Belgrade Serbia Shandong Provincial Key Laboratory of Computer Networks Shandong Fundamental Research Center for Computer Science Jinan China

With the rapid development of the new generation of information technologies, including cloud computing, 5G, and the Internet of Things, data outsourcing storage in the cloud has brought great convenience to data storage or access anytime and anywhere. Due to the separation of ownership and management of outsourced stored data, it faces data security risks, including data loss and personal privacy disclosure. The third-party audit of data operational behavior by logs is often employed to strengthen the security of outsourced data. However, some privacy breach incidents in recent years show that third parties who may leak information about data and users’ identities are not always trustworthy. To achieve data security and privacy protection, an operation behavior auditing scheme for outsourced data is proposed with the blockchain rather than third parties. The data operation behavior audit is carried out from three dimensions: integrity of the operation behavior log, the credibility of the operation behavior information, and compliance of operation behavior. The security analysis and experiment results indicate that this scheme is secure and feasible.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Adaboost-based Graph Convolutional networks with Channel Attention for Intrusion Detection in Internet-of-Things

Adaboost-based Graph Convolutional Networks with Channel Att...

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Jing Long Cuiting Luo Ruxin Chen Wei Liang Zheng Qin College of Information Science and Engineering Hunan Normal University Changsha China School of Computer Science and Engineering Hunan University of Science and Technology Xiangtan China College of Computer Science and Electronic Engineering Hunan University Changsha China

With the increasing proliferation of Internet of Things (IoT) and industrial big data technologies, while IoT scenarios offer intelligent services, cyberspace faces various types of threats and attacks. Despite Graph Convolutional networks (GCN) having been demonstrated as a potent tool for learning graph data representations, many current IoT intrusion detection methods tend to overlook the adoption of graph-related techniques, given the non-graph-structured nature of most IoT traffic data. This leads to the model design overlooking the potential correlations among variables, thus limiting the accuracy of intrusion detection. In this study, we introduce an integrated intrusion detection model that fuses graph learning with a dual attention mechanism(A-GCA). Initially, our research adopts a combination of GCN and Convolutional Neural networks (CNN) to deeply capture intricate interactions between data nodes. To further enhance the model’s feature-learning capability, we incorporate a channel attention mechanism for emphasizing features with pivotal information. Additionally, a multi-head attention mechanism is adopted to amalgamate a wealth of contextual information, ensuring more vivid feature representations. Regarding the training strategy, we employ a variant of Adaboost which uniquely steers weak classifiers from focusing on high-precision fitting of raw data towards precise binary classification tasks, aiming to augment the model’s overall performance in intrusion detection. Evaluations and performance analyses using real-world datasets UNSW-NB15 and CICIDS2017 indicate that the proposed method achieves improvements of 2.5% and 3.7% over the best baseline model in the accuracy and F1-score, respectively, relative to the baseline, surpassing other approaches.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Towards Energy-Aware Federated Traffic Prediction for Cellular networks

Towards Energy-Aware Federated Traffic Prediction for Cellul...

引用

international conference on Fog and Mobile Edge computing (FMEC)

作者： Vasileios Perifanis Nikolaos Pavlidis Selim F. Yilmaz Francesc Wilhelmi Elia Guerra Marco Miozzo Pavlos S. Efraimidis Paolo Dini Remous-Aris Koutsiamanis Department of Electrical and Computer Engineering Democritus University of Thrace Xanthi Greece Department of Electrical and Electronic Engineering Imperial College London London United Kingdom Radio Systems Research Nokia Bell Labs Stuttgart Germany Sustainable Artificial Intelligence Centre Tecnològic de Telecomunicacions de Catalunya (CTTC/CERCA) Barcelona Spain Department of Automation Production and Computer Sciences LS2N IMT Atlantique Inria Nantes France

Cellular traffic prediction is a crucial activity for optimizing networks in fifth-generation (5G) networks and beyond, as accurate forecasting is essential for intelligent network design, resource allocation and anomaly mitigation. Although machine learning (ML) is a promising approach to effectively predict network traffic, the centralization of massive data in a single data center raises issues regarding confidentiality, privacy and data transfer demands. To address these challenges, federated learning (FL) emerges as an appealing ML training framework which offers high accurate predictions through parallel distributed computations. However, the environmental impact of these methods is often overlooked, which calls into question their sustainability. In this paper, we address the trade-off between accuracy and energy consumption in FL by proposing a novel sustainability indicator that allows assessing the feasibility of ML models. Then, we comprehensively evaluate state-of-the-art deep learning (DL) architectures in a federated scenario using real-world measurements from base station (BS) sites in the area of Barcelona, Spain. Our findings indicate that larger ML models achieve marginally improved performance but have a significant environmental impact in terms of carbon footprint, which make them impractical for real-world applications.

关键词：

来源：评论

学校读者我要写书评

暂无评论

GP-GPU computing for Analysis of Large Scale networks 20

GP-GPU Computing for Analysis of Large Scale networks

引用

21st international conference on distributed computing and Networking (ICDCN)

作者： Ghoshal, Arnab Kumar Das, Nabanita Univ Calcutta Kolkata India Indian Stat Inst Kolkata India

ISBN: (纸本)9781450377515

Over last few years the interest in large scale network processing has gained momentum due to the increased importance of efficient processing of such networks for the analysis and problem solving in Social networks, Internet of Things(IoT), Data mining, Biological networks etc.. The steep increase in volume of data being produced, necessitates the development of parallel and distributed graph processing algorithms that demand extensive computation. Nowa days, because of high-computing-power of GP-GPUs (General Purpose Graphic Processor Unit) and its cost-effectiveness, GP-GPUs are being widely used in almost all areas of computing. In this paper, we address the problem of designing efficient parallel graph processing algorithms for GP-GPU platform which is the upcoming trend in high end machines. Presently, the problem of community detection and its influence on misinformation containment in online social networks have been investigated thoroughly, and parallel CUDA algorithms have been proposed. Simulation studies show a significant speed-up compared to the existing sequential/parallel techniques.

关键词： Cost effectiveness

来源：评论

学校读者我要写书评

暂无评论

WiRN: Real-Time and Lightweight Gesture Detection System on Edge Device 27

WiRN: Real-Time and Lightweight Gesture Detection System on ...

引用

27th IEEE international conference on parallel and distributed Systems, ICPADS 2021

作者： Yang, Qing Xing, Tianzhang Jiang, Zhiping Fu, Xinhua He, Jingyi School of Information Science and Technology Northwest University Xi'an China School Computer Science and Technology Xidian University Xi'an China

ISBN: (纸本)9781665408783

Gesture detection based on WiFi signals does not require users to carry additional equipment, and can better protect the privacy of users during the detection process, so it has received widespread attention. However, the existing work does not consider the actual deployment of the platform, and ignores the requirements for the computing power of the platform and the actual reasoning delay, resulting in many methods that are not suitable for the use of edge devices. In this paper, we propose a WiFi gesture detection system, named WiRN, which is fully deployed on edge devices and does not require the participation of additional computing devices. In WiRN, We have proposed solutions to related problems. First of all, in order to solve the problem of large differences in multiple phase differences obtained in different scenarios due to over-sensitive phases and to improve the robustness and universality of the system, we propose a multi-antenna-based phase difference selection algorithm to find the most suitable phase difference. Then, we fuse the amplitude and phase difference of different dimensions and obtain more fine-grained input data to solve the problem of the inability to deploy complex neural networks to fully extract features due to the limitation of edge device computing power, so that the input data contains richer feature information. In this way, for the first time, we will improve the accuracy of network classification from the data source. We evaluated the system through a series of experiments, and the results showed that under the premise of satisfying the real-time calculation of edge devices, we achieved the same accuracy as the existing complex network by using the simplest two-layer neural network. The recognition accuracy of about 93% is achieved in different environments. © 2021 IEEE.

关键词： Complex networks

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：