检索结果-内蒙古大学图书馆

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

作者： Xiang, Zhaohong Luo, Yigui Xie, Yin She, Haihong Liang, Weiming Zhao, Laigang Tongji University School of Software Engineering Shanghai China Company Limited China

ISBN: (纸本)9798350330717

Dynamic channel pruning is a technique aimed at reducing the theoretical computational complexity and inference latency of convolutional neural networks. Dynamic channel pruning methods introduce complex additional modules for dynamically selecting channels for images. Due to the additional modules, dynamic channel pruning methods never achieve optimal acceleration effect in real world. To address this problem, we propose Consecutive Dynamic Channel Pruning (CDCP), a novel dynamic channel pruning framework unified for almost all dynamic pruning methods designed for continuous image processing. The core idea of CDCP stems from our observation that adjusting the network for all frames in semantically continuous scenes is unnecessary since adjacent frames often share similar network structures in dynamic channel pruning. CDCP introduces a simple binary classifier to determine whether the network structure needs to be adjusted for a new frame. Our method can also be used for semantically non-continuous image processing tasks with a slightly lower probability of model reuse. We validate the effectiveness of CDCP on three dynamic channel pruning methods and better acceleration effects are achieved when applied them with CDCP to the semantically continuous Waymo dataset, the nuScenes dataset, and the semantically discontinuous COCO dataset. © 2023 IEEE.

关键词： dynamic channel pruning model acceleration network structure

来源：评论

学校读者我要写书评

暂无评论

distributed solar photovoltaic power prediction algorithm based on deep neural network

Journal of Engineering Research

引用

Journal of Engineering Research 2024年

作者： Ming Zhao Sitao Li Hao Chen Min Ling Hong Chang Digital Platform Technology Company China Southern Power Grid Shenzhen Guangdong 518000 China

Solar photovoltaic (PV) power prediction is easily affected by weather factors. In order to reduce the solar photovoltaic (PV) power prediction deviation and improve the prediction accuracy, a distributed solar photovoltaic (PV) power prediction algorithm based on deep neural network is proposed. By deeply exploring the working principle of photovoltaic power generation, constructing a photovoltaic power generation system model, and systematically analyzing various factors that affect photovoltaic power generation, detailed classification of weather types can be achieved. On this basis, outlier detection, standardization processing, and normalization techniques are used to deeply clean and optimize the raw data, effectively avoiding the problem of neuron saturation. The use of wavelet packet decomposition method to decompose the photovoltaic power generation sequence into multiple sub sequences significantly reduces the difficulty of prediction. The effective fusion of LSTM (Long Short-Term Memory) and BPNN (Back Propagation neural network), and the fine adjustment of the fusion ratio parameter through genetic algorithm, ultimately achieved high-precision prediction of distributed photovoltaic power under complex and variable weather conditions. The experimental results show that the proposed method can accurately predict photovoltaic power under different weather conditions, and the prediction results are reliable.

关键词： Deep neural network distributed photovoltaic Power prediction LSTM； BPNN； Genetic algorithm Weather factors

来源：评论

学校读者我要写书评

暂无评论

An Efficient Unstructured Sparse Convolutional neural network Accelerator for Wearable ECG Classification Device

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS 2022年第11期69卷 4572-4582页

作者： Lu, Jiahao Liu, Dongsheng Cheng, Xuan Wei, Lai Hu, Ang Zou, Xuecheng Huazhong Univ Sci & Technol Sch Opt & Elect Informat Wuhan 430074 Peoples R China

Convolution neural network (CNN) with pruning techniques has shown remarkable prospects in electrocardiogram (ECG) classification. However, efficiently deploying the existing pruned neural network to wearable devices for ECG classification is a great challenge due to the limited hardware resource and randomly distributed sparse weights. To address this issue, an efficient unstructured sparse CNN accelerator is proposed in this paper. A tile-first dataflow with compressed data storage format is presented to skip zero weight multiplications and increase the computing efficiency during inference of small-scale model with large sparsity. The two-level weight index matching structure in the dataflow exploits shifting operation to select valid data pairs and maintain the fully-pipelined calculation process. A configurable processing element (PE) array with 32-bit instruction control is proposed to increase the flexibility of the accelerator. Verified in FPGA and post-synthesis simulations in SMIC 40nm process, the proposed sparse CNN accelerator consumes 3.93 mu J/classification at 2MHz clock frequency and it achieves an averaged ECG classification accuracy of 98.99%. A computing efficiency of 118.75% is realized which is improved by 48% compared to the dense baseline. In brief, the proposed efficient CNN accelerator is especially suitable for wearable ECG classification device.

关键词： Wearable ECG classification unstructured sparse convolution neural network computing efficiency

来源：评论

学校读者我要写书评

暂无评论

Beamer: Stage-Aware Coflow Scheduling to Accelerate Hyper-Parameter Tuning in Deep Learning Clusters

引用

IEEE TRANSACTIONS ON network AND SERVICE MANAGEMENT 2022年第2期19卷 1083-1097页

作者： He, Yihong Cai, Weibo Zhou, Pan Sun, Gang Luo, Shouxi Yu, Hongfang Guizani, Mohsen Univ Elect Sci & Technol China Key Lab Opt Fiber Sensing & Commun Minist Educ Chengdu 611731 Peoples R China Agile & Intelligent Comp Key Lab Sichuan Prov Chengdu 611731 Peoples R China Southwest Jiaotong Univ Sch Comp & Artificial Intelligence Chengdu 611756 Peoples R China Peng Cheng Lab Shenzhen 518066 Peoples R China MBZUAI Machine Learning Dept Abu Dhabi U Arab Emirates

Training a neural network requires retraining the same model many times to search for the configuration of hyper-parameters with the best training result. It is common to launch multiple training jobs and evaluate them in stages. At the completion of each stage, jobs with unpromising configurations will be terminated and jobs with new configurations will start. Each job typically performs distributed training across multiple GPUs, and GPUs periodically synchronize their models over the network. However, model synchronizations of running jobs cause severe network congestion, significantly increasing the stage completion time (SCT) and thus the time to successfully search for the desired configuration. Existing flow schedulers are ineffective to reduce SCT since they are agnostic to training stages. In this paper, we propose a stage-aware coflow scheduling method to minimize the average SCT. In this method, an algorithm is designed to order coflows by considering stage information and then coflows are scheduled according to the order. Mathematical analysis shows that the method achieves the average SCT within 20/3 of the optimal. We implement the method in a real system called Beamer. Extensive testbed experiments and simulations show that Beamer significantly outperforms advanced network designs, such as Sincronia, FIFO-LM, and per-flow fair sharing.

关键词： Training Data models Deep learning Servers Schedules Graphics processing units Tuning Deep learning flow scheduling hyper-parameter tuning search neural network stage completion time

来源：评论

学校读者我要写书评

暂无评论

Application of neural networks in Data Communication Analysis 12

Application of Neural Networks in Data Communication Analysi...

引用

12th International Scientific Conference on Communication and Information Technologies, KIT 2023

作者： Badidova, Bianca Turcanik, Michal Armed Forces Academy of General M. R.Štefánik Department of Informatics Liptovský Mikuláš Slovakia

ISBN: (纸本)9798350338393

The purpose of this paper is to analyze modern trends in the field of data communication analysis with a specific focus on identifying commonly used methodologies in the detection of unusual communication patterns. This paper provides a brief overview of the techniques employed in data communication analysis. We have examined particular techniques used for identifying abnormal communication patterns within network packets, associated with malicious attacks and their features. Given the information, we analyzed the utilization of artificial intelligence techniques for identifying unusual communication patterns, using feedforward neural network topologies. We proposed two types of topologies to address this problem - multilayer perceptron and convolutional neural network. The study includes the creation of unique datasets that contain both our generated data and openly available datasets, which are subsequently analyzed. The outcomes of this research include an assessment of the classification capabilities, solution complexity and processing time of multilayer perceptron and convolutional neural network topologies used in data communication analysis, based on the specific data flow and used computational system. © 2023 IEEE.

关键词： Multilayer neural networks

来源：评论

学校读者我要写书评

暂无评论

Multiple Sound Source Localization in Three Dimensions Using Convolutional neural networks and Clustering Based Post-processing

引用

IEEE ACCESS 2022年 10卷 125707-125722页

作者： Sakavicius, Saulius Serackis, Arturas Abromavicius, Vytautas Vilnius Tech Fac Elect Dept Elect Syst Vilnius 10221 Lithuania

Sound source localization methods are successfully applied for various estimation tasks, such as tracking and detecting objects, aiming cameras, and navigating robots. However, large and usually complex distributed microphone arrays are used for three-dimensional acoustic source localization. This study proposes a convolutional neural network architecture for three-dimensional sound source localization using a single tetrahedral microphone array. A spectrum phase component of a microphone array signal was designed as the input of the model, while the output represents a three-dimensional space. The paper provides extensive experimental results of the given method on a semi-synthetic audio data set and a real-world microphone array. Furthermore, cluster-based post-processing has been shown to increase the accuracy of three-dimensional localization by more than 30%. The experimental results on a synthesized data set using the image source method showed 1.08 m localization uncertainties. The estimate of the investigated sound sources had a mean absolute error of 18.97 & DEG;and elevation error of 48.49 & DEG;. An additional advantage of the proposed method is the ability to predict the location of the sound source from a single signal analysis frame. This gives instant localization and is in line with many alternative applications. The proposed solution does not require intensive preprocessing of the audio signals and can be used as a video camera pointing system based on a microphone array. In the future, it would be relevant to investigate the localization performance of more than two sound sources, and the variable acoustic conditions could also be assessed.

关键词： Location awareness Microphone arrays Acoustics Clustering algorithms Direction-of-arrival estimation Cameras Estimation Clustering convolutional neural networks microphone array multiple sound sources sound source localization three-dimensional

来源：评论

学校读者我要写书评

暂无评论

An Accelerator for Sparse Convolutional neural networks Leveraging Systolic General Matrix-matrix Multiplication

引用

ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION 2022年第3期19卷 1–26页

作者： Soltaniyeh, Mohammadreza Martin, Richard P. Nagarakatte, Santosh Rutgers State Univ 110 Frelinghuysen Rd SAS CSBusch Campus Piscataway NJ 08854 USA

This article proposes a novel hardware accelerator for the inference task with sparse convolutional neural networks (CNNs) by building a hardware unit to perform Image to Column (IM2COL) transformation of the input feature map coupled with a systolic-array-based general matrix-matrix multiplication (GEMM) unit. Our design carefully overlaps the IM2COL transformation with the GEMM computation to maximize parallelism. We propose a novel design for the IM2COL unit that uses a set of distributed local memories connected by a ring network, which improves energy efficiency and latency by streaming the input feature map only once. The systolic-array-based GEMM unit in the accelerator can be dynamically configured as multiple GEMM units with square-shaped systolic arrays or as a single GEMM unit with a tall systolic array. This dynamic reconfigurability enables effective pipelining of IM2COL and GEMM operations and attains high processing element utilization for a wide range of CNNs. Further, our accelerator is sparsity aware, improving performance and energy efficiency by effectively mapping the sparse feature maps and weights to the processing elements, skipping ineffectual operations and unnecessary data movements involving zeros. Our prototype, SPOTS, is on average 2.16x, 1.74x, and 1.63x faster than Gemmini, Eyeriss, and Sparse-PE, which are prior hardware accelerators for dense and sparse CNNs, respectively. SPOTS is also 78x and 12x more energy-efficient when compared to CPU and GPU implementations, respectively.

关键词： Convolutional neural networks hardware accelerators systolic arrays GEMM sparse computation

来源：评论

学校读者我要写书评

暂无评论

Graph neural networks With Convolutional ARMA Filters

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2022年第7期44卷 3496-3507页

作者： Bianchi, Filippo Maria Grattarola, Daniele Livi, Lorenzo Alippi, Cesare UiT Arctic Univ Norway Dept Math & Stat N-9019 Tromso Norway NORCE Norwegian Res Ctr N-5008 Bergen Norway Univ Svizzera Italiana Fac Informat CH-6900 Lugano Switzerland Univ Manitoba Dept Comp Sci & Math Winnipeg MB R3T 2N2 Canada Univ Exeter Dept Comp Sci Exeter EX4 4PY Devon England Politecn Milan Dept Elect Informat & Bioengn I-20133 Milan Italy

Popular graph neural networks implement convolution operations on graphs based on polynomial spectral filters. In this paper, we propose a novel graph convolutional layer inspired by the auto-regressive moving average (ARMA) filter that, compared to polynomial ones, provides a more flexible frequency response, is more robust to noise, and better captures the global graph structure. We propose a graph neural network implementation of the ARMA filter with a recursive and distributed formulation, obtaining a convolutional layer that is efficient to train, localized in the node space, and can be transferred to new graphs at test time. We perform a spectral analysis to study the filtering effect of the proposed ARMA layer and report experiments on four downstream tasks: semi-supervised node classification, graph signal classification, graph classification, and graph regression. Results show that the proposed ARMA layer brings significant improvements over graph neural networks based on polynomial filters.

关键词： Convolution Laplace equations Task analysis Graph neural networks Chebyshev approximation Frequency response Eigenvalues and eigenfunctions Geometric deep learning graph filters graph neural networks graph theory graph signal processing

来源：评论

学校读者我要写书评

暂无评论

A Model for Vulnerability Classification based on Res-CNN-BiTLSTM 22

A Model for Vulnerability Classification based on Res-CNN-Bi...

引用

22nd IEEE International Symposium on Parallel and distributed processing with Applications, ISPA 2024

作者： Li, Jiazheng Ren, Jiadong Yang, Shuailin Zhi, Chenghao Bao, Chunjiao Yanshan University School of Information Science and Engineering Qinhuangdao China

ISBN: (纸本)9798331509712

Vulnerabilities in software and distributed systems are increasing, and system security becomes a challenge for developers. Giving a quick vulnerability classification for newly discovered vulnerabilities is helpful for developers to quickly analyze the vulnerabilities and complete the vulnerability fix. Therefore, Res-CNN-BiTLSTM is proposed for vulnerability classification in this paper. The model consists of Text Convolutional neural network (TextCNN), Bidirectional improved gating mechanism (BiTLSTM) and residual blocks. Firstly, TextCNN is employed to capture local features from vulnerability description information, according to different convolutional kernel sizes. Secondly, an improved gating mechanism (TLSTM) is constructed to enhance the expressive power of Long Short-Term Memory (LSTM). BiTLSTM is employed to extract long-text features of vulnerability descriptions through both forward and backward directions. Thirdly, the residual block is used for enhancing and fusing the local features with the long text features to retain important text features. Finally, a fully connected layer is employed to classify vulnerabilities. The results show that Res-CNN-BiTLSTM has the highest macro precision (MacroP), macro recall (MacroR), and macro F1 (MacroF1) on National Vulnerability Database (NVD) dataset. It has a high prediction precision on CWE-352 and CWE-416. The performance of Res-CNN-BiTLSTM is better than the other comparative models. It is effective for vulnerability automatic classification. Meanwhile, the effectiveness of each part of the model is proved by ablation experiments. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Optimization and acceleration of convolutional neural networks: A survey

引用

JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES 2022年第7期34卷 4244-4268页

作者： Habib, Gousia Qureshi, Shaima Natl Inst Technol Srinagar Dept Comp Sci & Engn Srinagar Jammu & Kashmir India

Convolutional neural networks (CNN) is a specialized case of artificial neural networks(ANN) and finds its application in computer vision and parallel distributed computing for processing of massive amount of data generated by sensors and to meet the power constraints of IOT devices. Recent advancements in parameter optimization, regularization techniques, improvement in activation functions, corresponding loss functions, advancements in the coted the research of Convolutonal neural network's(CNN's) in past few years. Training of neural networks is cumbersome and takes a lot of time can take days or even weeks. This limits the application of Convolutional neural network(CNN) in real time research fields where computational speed is of utmost importance. Thus there is a need for appropriate and enhanced computational speed to meet the requirements of these real time applications. This paper describes CNN in detail summarizes architectural evolution of CNN from 1998 to 2019. Three types of strategies have been explained to enhance the computational speed of CNN at algorithmic level and implementation level. This paper gives detailed insight about computation speed acceleration using Stochastic Gradient Decent(SGD) optimization, Fast convolution and exploiting parallelism challenges in CNN posed by these techniques and recent advancements. The paper also includes detailed view of different framework usage while implementing fast convolution or parallelism techniques. The ultimate aim of the paper to explore all such recent techniques by which we can accelerate the training speed of the CNN's without compromising the accuracy. (C) 2020 The Authors. Published by Elsevier B.V. on behalf of King Saud University.

关键词： SGD ReLu HyPar ADAM NADAM

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：