检索结果-内蒙古大学图书馆

37th IEEE International Parallel and distributed processing Symposium, IPDPS 2023

ISBN: (纸本)9798350337662

The proceedings contain 95 papers. The topics discussed include: distributed sparse random projection trees for constructing K-nearest neighbor graphs;fast deterministic gathering with detection on arbitrary graphs: the power of many robots;accurate and efficient distributed covid-19 spread prediction based on a large-scale time-varying people mobility graph;accelerating packet processing in container overlay networks via packet-level parallelism;efficient hardware primitives for immediate memory reclamation in optimistic data structures;efficient hardware primitives for immediate memory reclamation in optimistic data structures;accelerating distributed deep learning training with compression assisted Allgather and reduce-scatter communication;accelerating CNN inference on long vector architectures via co-design;exploiting input tensor dynamics in activation checkpointing for efficient training on GPU;drill: log-based anomaly detection for large-scale storage systems using source code analysis;dynasparse: accelerating GNN inference through dynamic sparsity exploitation;exploiting sparsity in pruned neural networks to optimize large model training;SRC: mitigate I/O throughput degradation in network congestion control of disaggregated storage systems;boosting multi-block repair in cloud storage systems with wide-stripe erasure coding;on doorway egress by autonomous robots;and on the arithmetic intensity of distributed-memory dense matrix multiplication involving a symmetric input matrix (SYMM).

关键词：

来源：评论

学校读者我要写书评

暂无评论

Gradient and self-attention enabled convolutional neural network for crack detection in smart cities 29

Gradient and self-attention enabled convolutional neural net...

引用

29th IEEE International Conference on Parallel and distributed Systems, ICPADS 2023

作者： Xie, Renping Chen, Mengyao Tao, Ming Ding, Kai Chen, HaoHan Dongguan University of Technology Dongguan China

ISBN: (纸本)9798350330717

Intelligent transportation is an important guarantee for the safety and efficiency of urban transportation in smart cities, and regular road pavement inspection is the focus of road and bridge maintenance in intelligent transportation. Cracks in concrete pavement are the most common type of pavement damage and the earliest sign of pavement deterioration. However, existing crack detection algorithms suffer from incomplete crack detection and are easily disturbed by pseudo-cracks such as water spots and leaves. To address the above problems, this paper proposes a convolutional neural network (CNN) method that introduces a gradient module and an attention mechanism. The method adopts a CNN model based on the VGG-16 structure as the main body of the network structure, and optimally adjusts the network structure by incorporating a gradient layer and a self-attention mechanism, accelerating the convergence speed of network training and the global information learning ability. A negative sample dataset with pseudo-cracks, such as leaves, water spots and branches was constructed, and comparative experimental analysis was conducted in terms of both visual judgment and objective indicators. The experimental results show that after the introduction of the gradient layer and the self-attentive mechanism, not only the convergence speed of the network training is faster, but also the cracks in the concrete pavement images can be segmented more completely and accurately. © 2023 IEEE.

关键词： convolution neural network crack detection image segmentation

来源：评论

学校读者我要写书评

暂无评论

Survey and Enhancements on Deploying LSTM Recurrent neural networks on Embedded Systems

Survey and Enhancements on Deploying LSTM Recurrent Neural N...

引用

IEEE International Conference on Communications (IEEE ICC)

作者： Abib, Ghalid Castel, Florian Satouri, Nissrine Afifi, Hossam Said, Adel Mounir Inst Polytech Paris SAMOVAR Telecom SudParis Palaiseau France Inst Polytech Paris Telecom SudParis Palaiseau France Natl Telecommun Inst NTI Switching Dept Cairo Egypt

ISBN: (纸本)9781538674628

The real implementation of a recurrent neural network (RNN) in a low complexity IoT device is evaluated in order to predict the time series of power consumption in tertiary buildings. The RNN type long short-term memory (LSTM) algorithm is adapted for a 32-bit microcontroller unit (MCU) and the backpropagation (BP) algorithm is implemented in-house. We therefore demonstrate that Intelligent IoT (IIoT) devices, such as the Espressif ESP32 MCU, not only implement neural networks (NNs), but also learn on their own. The resulting IIoT architecture has been proven to operate efficiently and compared to the traditional computer-based learning platform. The selected results confirm that stand-alone IoT devices are a truly efficient solution that adds flexibility to the architecture, reduces storage and computation costs, and is more energy-friendly. As a conclusion, it is practically more efficient to exploit low-power and processing-time IIoT for our prediction use case rather than relying on server based distributed systems.

关键词： Intelligent IoT Machine Learning Recurrent neural networks Long Short-Term Memory Edge AI Embedded Systems

来源：评论

学校读者我要写书评

暂无评论

GraphTheta: A distributed Graph neural network Learning System With Flexible Training Strategy

arXiv

引用

arXiv 2021年

作者： Liu, Yongchao Li, Houyi Zhang, Guowei Zeng, Xintan Li, Yongyong Huang, Bin Zhang, Peng Li, Zhao Zhu, Xiaowei He, Changhua Chen, Wenguang Ant Group China Fudan University China Guangzhou University China Zhejiang University China Tsinghua University China

Graph neural networks (GNNs) have been demonstrated as a powerful tool for analyzing non-Euclidean graph data. However, the lack of efficient distributed graph learning systems severely hinders applications of GNNs, especially when graphs are big and GNNs are relatively deep. Herein, we present GraphTheta, the first distributed and scalable graph learning system built upon vertex-centric distributed graph processing with neural network operators implemented as user-defined functions. This system supports multiple training strategies and enables efficient and scalable big-graph learning on distributed (virtual) machines with low memory. To facilitate graph convolutions, GraphTheta puts forward a new graph learning abstraction named NN-TGAR to bridge the gap between graph processing and graph deep learning. A distributed graph engine is proposed to conduct the stochastic gradient descent optimization with a hybrid-parallel execution, and a new cluster-batched training strategy is supported. We evaluate GraphTheta using several datasets with network sizes ranging from small-, modest- to large-scale. Experimental results show that GraphTheta can scale well to 1,024 workers for training an in-house developed GNN on an industry-scale Alipay dataset of 1.4 billion nodes and 4.1 billion attributed edges, with a cluster of CPU virtual machines (dockers) of small memory each (5∼12GB). Moreover, GraphTheta can outperform DistDGL by up to 2.02×, with better scalability, and GraphLearn by up to 30.56×. As for model accuracy, GraphTheta is capable of learning as good GNNs as existing frameworks. To the best of our knowledge, this work presents the largest edge-attributed GNN learning task in the literature. Copyright © 2021, The Authors. All rights reserved.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

A Learning-Based Scheduler for High Volume processing in Data Warehouse Using Graph neural networks 22nd

A Learning-Based Scheduler for High Volume Processing in Dat...

引用

22nd International Conference on Parallel and distributed Computing, Applications and Technologies (PDCAT 2021)

作者： Bengre, Vivek HoseinyFarahabady, M. Reza Pivezhandi, Mohammad Zomaya, Albert Y. Jannesari, Ali Iowa State Univ Dept Comp Sci Lab Software Analyt & Pervas Parallelism SwAPP Ames IA USA Univ Sydney Ctr Distributed & High Performance Comp Sch Comp Sci Camperdown NSW Australia

ISBN: (纸本)9783030967727;9783030967710

The process of extracting, transforming, and loading (also known as ETL) of a high volume of data plays an essential role in data integration strategies in data warehouse systems in recent years. In almost all distributed ETL systems currently use in both industrial and academia context, a simple heuristic-based scheduling policy is employed. Such a heuristic policy tries to process a stream of jobs in the best-effort fashion, however, it can result in under-utilization of computing resources in most practical scenarios. On the other hand, such inefficient resource allocation strategy can result in an unwanted increase in the total completion time of data processing jobs. In this paper, we develop an efficient reinforcement learning technique that uses a Graph neural network (GNN) model to combine all submitted tasks graphs into a single graph to simplify the representation of the states within the environment and efficiently make a parallel application for processing of the submitted jobs. Besides, to positively augment the embedding features in each leaf node, we pass messages from leaf to root so the nodes can collaboratively represent actions within the environment. The performance results show up to 15% improvement in job completion time compared to the state-of-the-art machine learning scheduler and up to 20% enhancement compared to a tuned heuristic-based scheduler.

关键词： Extract Transform Load (ETL) operations Scheduling policy Data streaming processing system Graph neural networks Job completion time Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

FACVSPO: Fractional anti corona virus student psychology optimization enabled deep residual network and hybrid correlative feature selection for distributed denial-of-service attack detection in cloud using spark architecture

引用

INTERNATIONAL JOURNAL OF ADAPTIVE CONTROL AND SIGNAL processing 2022年第7期36卷 1647-1669页

作者： Ramkumar, Muthuperumal Periyaperumal Ramasamy, Ganeshan Azees, Maria Sakthi, Ulaganathan Thiagarajar Coll Engn Dept Comp Sci & Engn Madurai 625015 Tamil Nadu India VIT Bhopal Univ Sch Comp Sci & Engn Bhopal Madhya Pradesh India GMR Inst Technol Dept Elect & Commun Engn Rajam Andra Pradesh India SIMATS Saveetha Sch Engn Dept Comp Sci & Engn Chennai Tamil Nadu India

Cloud computing is an emerging standard in modern days for the purpose of sharing huge data, as it affords numerous user friendly behaviors. Cloud computing services offer an extensive range of resource pool in order to maintain huge scale data. Although, cloud computing model is disposed to several cyber-attacks and security problems regarding cloud structure, because of the dynamic and distribute character and exposures in virtualization implementation. distributed denial-of-service (DDoS) attack is a type of cyber-attack, which disturbs the usual traffic of targeted cloud server. Moreover, DDoS produces malicious traffic in cloud structure, and thus consumes cloud resources. In this paper, an effective DDoS attack detection model, named fractional anti corona virus student psychology optimization-based deep residual network (FACVSPO-based DRN) is implemented using spark architecture. The devised FACVSPO approach is newly designed by incorporating anti coronavirus optimization (ACVO) algorithm, fractional calculus (FC) and student psychology based optimization (SPBO) model. Moreover, the hybrid correlative scheme is designed for extracting significant features for attack detection. The DRN structure is utilized for performing attack recognition, which categorizes the data as normal or attack. In addition, the DRN classifier is trained by the developed FACVSPO approach. The developed attack detection model outperformed other existing techniques in terms of testing accuracy, true negative rate (TNR), true positive rate (TPR) of 0.9236, 0.9141, and 0.9412, respectively. The testing accuracy of the implemented model is 12.02%, 8.92%, 7.27%, 6.30%, 5.68%, and 1.20% better than the existing methods, such as Taylor-elephant herd optimisation based deep belief network (TEHO-DBN), deep learning, deep neural network (DNN), multiple kernel learning, Fuzzy Taylor elephant herd optimisation (EHO)-based DBN, fractional anti corona virus optimization-deep neuro fuzzy network (FA

关键词： anti coronavirus optimization algorithm deep residual network distributed denial-of-service fractional calculus student psychology based optimization model

来源：评论

学校读者我要写书评

暂无评论

Accelerating Large Sparse neural network Inference Using GPU Task Graph Parallelism

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2022年第11期33卷 3041-3052页

作者： Lin, Dian-Lun Huang, Tsung-Wei Univ Utah Dept Elect & Comp Engn Salt Lake City UT 84112 USA

The ever-increasing size of modern deep neural network (DNN) architectures has put increasing strain on the hardware needed to implement them. Sparsified DNNs can greatly reduce memory costs and increase throughput over standard DNNs, if the loss of accuracy can be adequately controlled. However, sparse DNNs present unique computational challenges. Efficient model or data parallelism algorithms are extremely hard to design and implement. The recent effort MIT/IEEE/Amazon HPEC Graph Challenge has drawn attention to high-performance inference methods for large sparse DNNs. In this article, we introduce SNIG, an efficient inference engine for large sparse DNNs. SNIG develops highly optimized inference kernels and leverages the power of CUDA Graphs to enable efficient decomposition of model and data parallelisms. Our decomposition strategy is flexible and scalable to different partitions of data volumes, model sizes, and GPU numbers. We have evaluated SNIG on the official benchmarks of HPEC Sparse DNN Challenge and demonstrated its promising performance scalable from a single GPU to multiple GPUs. Compared to the champion of the 2019 HPEC Sparse DNN Challenge, SNIG can finish all inference workloads using only a single GPU. At the largest DNN, which has more than 4 billion parameters across 1920 layers each of 65536 neurons, SNIG is up to 2.3x faster than a state-of-the-art baseline under a machine of 4 GPUs. SNIG receives the Champion Award in 2020 HPEC Sparse DNN Challenge.

关键词： Graphics processing units Kernel Task analysis Parallel processing Programming Neurons Data models Task graph parallelism

来源：评论

学校读者我要写书评

暂无评论

DVPPIR: privacy-preserving image retrieval based on DCNN and VHE

引用

neural COMPUTING & APPLICATIONS 2022年第17期34卷 14355-14371页

作者： Li, Su Wu, Lei Meng, Weizhi Xu, Zihui Qin, Chengyi Wang, Hao Shandong Normal Univ Sch Informat Sci & Engn Jinan Peoples R China Henan Key Lab Network Cryptog Technol Zhengzhou Peoples R China Shandong Prov Key Lab Novel Distributed Comp Soft Jinan Peoples R China Tech Univ Denmark Dept Appl Math & Comp Sci Lyngby Denmark

With 5G and Internet technologies developing rapidly, outsourcing images to cloud servers has attracted growing attention. In existing technologies, images are often outsourced to cloud servers to reduce storage and computing burdens. However, outsourcing images to cloud servers without any processing may reveal the users' privacy, because the images may contain sensitive information about users, such as faces and locations, especially in electronic investigation. To overcome the security problems in image retrieval, we propose a privacy-preserving image retrieval scheme based on deep convolutional neural network (DCNN) and vector homomorphic encryption (VHE). We adopt DCNN and hash algorithms to extract image feature vectors, which improves retrieval accuracy. By combining VHE and K-means outsourcing clustering algorithms, the cloud server can build encrypted index trees, which speeds up the search and reduces the computational cost. In addition, a lightweight access control technique is used to allow image owners to set access policies for datasets flexibly. We prove the security of the proposed scheme and show the effectiveness of the scheme through experiments. Our scheme is suitable for application in electronic image investigation systems (EIIs) to optimize the storage and search of police data.

关键词： Privacy-preserving EIIs DCNN VHE K-means outsourcing Access control

来源：评论

学校读者我要写书评

暂无评论

Design and Application of Web-based Education Management System using Split and Kernel based Residual network

Design and Application of Web-based Education Management Sys...

引用

2024 International Conference on Intelligent Algorithms for Computational Intelligence Systems, IACIS 2024

作者： Wang, Yan South West Minzu University School of Public Adminisration Chengdu China

ISBN: (纸本)9798350360660

A web based education management system is established to develop education system by enhancing quality of education and teaching model. However, the existing resource allocation model and teaching in web-based education system has limitations such as ineffective teaching due to inappropriate resource allocation affect the education system. To overcome this problem, a Split and Kernel based Residual network (SK-ResNet) is proposed to improve quality of education by allocating resources effectively and enhance teaching model in the web based education system. The split technique in ResNet model, divide the tasks and distributed across multiple servers or devices. By splitting and kernel optimizing the tasks, the load on any single system allows the model for faster content processing and delivery. At first, the data are acquired and fed into functional structure of education system, and finally analyzed and integrated into cloud computing based on SK-ResNet model. The experimental results of the proposed method achieved accuracy of 87.29% which is greater when compared to existing method such as Deep neural network (DNN) and Hybrid Deep Learning (HDL) model. © 2024 IEEE.

关键词： Cloud platforms

来源：评论

学校读者我要写书评

暂无评论

Advanced scene text recognition and application for dynamic driving scenes 4

Advanced scene text recognition and application for dynamic ...

引用

4th International Conference on Signal Image processing and Communication, ICSIPC 2024

作者： Wang, Jiahao Lu, Yahao Wu, Lianpei School of Information Engineering Guangdong University of Technology Guangzhou510006 China

ISBN: (纸本)9781510682467

In the mobile driving scenario, insufficient data has become a major challenge for the application of scene text recognition models. An alternative to reduce the cost of data annotation is the active learning method, which improves the performance of the model by screening the data with the largest annotation information entropy for training. However, the calibration deviation of modern convolutional neural network is large, and the confidence cannot accurately reflect the real situation of model prediction. In view of the above problems, a scene text recognition framework based on active learning is proposed. The strategy of generating identically distributed heterogeneous data based on anti-aliasing operation is introduced into the framework. A confidence evaluation method based on prediction invariance is proposed. Combined with the active learning method, the confidence of sample prediction is evaluated and corrected. In addition, a text recognition dataset for mobile driving scenarios is established. This method tested SAR, SVTR, and RobustScanner models on the DSO dataset, with accuracy improvements of 11.28%, 15.12%, and 11.66%, respectively. Compared with experiments with randomly annotated data, the accuracy gains of each model were 5.32%, 5.88%, and 6.53%, respectively. The results confirm that this method significantly reduces manual annotation costs while enhancing model performance and robustness. © 2024 SPIE.

关键词： Convolutional neural 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：