检索结果-南通市图书馆

维普期刊数据库

Fast Fourier transform convolution neural network accelerator based on overlap addition

在线全文

学校读者我要写书评

暂无评论

The Journal of China Universities of Posts and Telecommunications 2024年

作者： You Chen Li Dejian Feng Xi Shen Chongfei Wei Jizeng Liu Yu School of Microelectronics Tianjin University College of intelligence and Computing Tianjin University Beijing Smart-Chip Microelectronics Technology company limited

In convolutional neural networks（CNNs）， the floating point computation in the traditional convolution layer is enormous， and the execution speed of the network is limited by intensive computing， which makes it challenging to meet the real-time response requirements of complex applications. This work is based on the principle that the time domain convolution result equals the frequency domain dot multiplication result to reduce the amount of floating-point calculations for convolution. The input feature map and the convolution kernel are converted to the frequency domain by the fast Fourier transform（FFT）， and the corresponding dot multiplication is performed. Then the frequency domain result is converted back to the time domain， and the output result of the convolution is obtained. In the shared CNN， the input feature map is much larger than the convolution kernel， resulting in many invalid operations. The overlap addition method is proposed to reduce invalid calculations and speed up network execution better. This work designs a hardware accelerator for frequency domain convolution and verifies its efficiency on the Xilinx ZYNQ UltraScale+MPSoC ZCU102 board: comparing the calculation time of Visual Geometry Group16(VGG16) under the ImageNet dataset， compared to traditional convolution accelerators， the hardware acceleration of frequency-domain convolution has improved performance by 8.5 times.

关键词：

Research and design of an expert diagnosis system for rail vehicle driven by data mechanism models

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Railway Sciences 2024年第4期3卷 480-502页

作者： Lin Li Jiushan Wang Shilu Xiao Zhuzhou Guochuang Rail Technology company limited ZhuzhouChina Zhuzhou Guochuang Rail Technology company limited Industrial Intelligence Research InstituteZhuzhouChina

Purpose-The aim of this work is to research and design an expert diagnosis system for rail vehicle driven by data mechanism ***/methodology/approach-The expert diagnosis system utilizes statistical and deep learning methods to model the real-time status and historical data features of rail *** on data mechanism models,it predicts the lifespan of key components,evaluates the health status of the vehicle and achieves intelligent monitoring and diagnosis of rail ***-The actual operation effect of this system shows that it has improved the intelligent level of the rail vehicle monitoring system,which helps operators to monitor the operation of vehicle online,predict potential risks and faults of vehicle and ensure the smooth and safe operation of ***/value-This system improves the efficiency of rail vehicle operation,scheduling and maintenance through intelligent monitoring and diagnosis of rail vehicle.

关键词： Rail transit Rail vehicle Expert diagnosis Intelligent operation and maintenance Deep learning Lifespan prediction Reliability analysis

Privacy-Preserving Frank-Wolfe on Shuffle Model

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Acta Mathematicae Applicatae Sinica 2024年

作者： Ling-jie ZHANG Shi-song WU Hai ZHANG School of Mathematics Northwest University School of Mathematics and Information Science Baoji University of Arts and Sciences China Southern Power Grid Artificial intelligence Technology company limited

In this paper,we design the differentially private variants of the classical Frank-Wolfe algorithm with shuffle model in the optimization of machine *** weak assumptions and the generalized linear loss（GLL） structure,we propose a noisy Frank-Wolfe with shuffle model algorithm（NoisyFWS） and a noisy variance-reduced Frank-Wolfe with the shuffle model algorithm（NoisyVRFWS） by adding calibrated laplace noise under shuffling scheme in the lp（p∈ [1,2」）-case,and study their privacy as well as utility guarantees for the H?lder smoothness *** particular,the privacy guarantees are mainly achieved by using advanced composition and privacy amplification by *** utility bounds of the NoisyFWS and NoisyVRFWS are analyzed and obtained the optimal excess population risks O(n-（1+a）/4a+log（d）√log（1/δ）/nε) and O(n-（1+a）/4a+log（d）√log（1/δ）/n2ε) with gradient complexity O(n（1+α）2/4α2) for α∈[1/√3,1].It turns out that the risk rates under shuffling scheme are a nearly-dimension independent rate,which is consistent with the previous work in some *** addition,there is a vital tradeoff between（α,L）-H?lder smoothness GLL and the gradient *** linear gradient complexity O（n） is showed by the parameter α=1.

关键词：

Early warning of core network capacity in space-terrestrial integrated networks

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Journal of Systems Engineering and Electronics 2024年第4期35卷 855-864页

作者： HAN Sai LI Ao ZHANG Dongyue ZHU Bin WANG Zelin WANG Guangquan MIAO Jie MA Hongbing China Unicom Research Institute Beijing 100044China China United Network Communications Group company limited Beijing 100033China

With the rapid development of low-orbit satellite com-munication networks both domestically and internationally,space-terrestrial integrated networks will become the future development *** space and terrestrial networks with limi-ted resources,the utilization efficiency of the entire space-terres-trial integrated networks resources can be affected by the core network *** order to improve the response efficiency of core networks expansion construction,early warning of the core network elements capacity is *** on the inte-grated architecture of space and terrestrial network,multidimen-sional factors are considered in this paper,including the number of terminals,login users,and the rules of users’migration during *** artifical intelligence(AI)technologies,the regis-tered users of the access and mobility management function(AMF),authorization users of the unified data management(UDM),protocol data unit(PDU)sessions of session manage-ment function(SMF)are predicted in combination with the num-ber of login users,the number of ***,the core network elements capacity can be predicted in *** proposed method is proven to be effective based on the data from real network.

关键词： space-terrestrial integrated networks core network element capacity artificial intelligent(AI)

维普期刊数据库

Three-dimensional diabetic macular edema thickness maps based on fluid segmentation and fovea detection using deep learning

在线全文

学校读者我要写书评

暂无评论

International Journal of Ophthalmology(English edition) 2022年第3期15卷 495-501页

作者： Jing-Jing Xu Yang Zhou Qi-Jie Wei Kang Li Zhen-Ping Li Tian Yu Jian-Chun Zhao Da-Yong Ding Xi-Rong Li Guang-Zhi Wang Hong Dai School of Medicine Tsinghua UniversityBeijing 100084China visionary intelligence company limited Beijing 100872China Department of Ophthalmology Beijing HospitalNational Center of GerontologyInstitute of Geriatric MedicineChinese Academy of Medical ScienciesBeijing 100730China Key Lab of Data Engineering and Knowledge Engineering Renmin University of ChinaBeijing 100872China

AIM: To explore a more accurate quantifying diagnosis method of diabetic macular edema(DME) by displaying detailed 3D morphometry beyond the gold-standard quantification indicator-central retinal thickness(CRT) and apply it in follow-up of DME ***: Optical coherence tomography(OCT) scans of 229 eyes from 160 patients were *** manually annotated cystoid macular edema(CME), subretinal fluid(SRF) and fovea as ground *** convolution neural networks(DCNNs) were constructed including U-Net, sASPP, HRNetV2-W48, and HRNetV2-W48+Object-Contextual Representation(OCR) for fluid(CME+SRF) segmentation and fovea detection respectively, based on which the thickness maps of CME, SRF and retina were generated and divided by Early Treatment Diabetic Retinopathy Study(ETDRS) ***: In fluid segmentation, with the best DCNN constructed and loss function, the dice similarity coefficients(DSC) of segmentation reached 0.78(CME), 0.82(SRF), and 0.95(retina).In fovea detection, the average deviation between the predicted fovea and the ground truth reached 145.7±117.8 μ*** generated macular edema thickness maps are able to discover center-involved DME by intuitive morphometry and fluid volume, which is ignored by the traditional definition of CRT>250 μ*** maps could also help to discover fluid above or below the fovea center ignored or underestimated by a single OCT ***: Compared to the traditional unidimensional indicator-CRT, 3D macular edema thickness maps are able to display more intuitive morphometry and detailed statistics of DME, supporting more accurate diagnoses and follow-up of DME patients.

关键词： diabetic macular edema fluid segmentation fovea detection 3D macular edema thickness maps deep learning

Unsupervised Dialogue State Tracking for End-to-End Task-Oriented Dialogue with a Multi-Span Prediction Network

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Computer Science & Technology 2023年第4期38卷 834-852页

作者：刘庆斌何世柱刘操刘康赵军 National Laboratory of Pattern Recognition Institute of AutomationChinese Academy of SciencesBeijing 100190China School of Artificial intelligence University of Chinese Academy of SciencesBeijing 100049China Beijing Sankuai Online Technology company limited Beijing 100102China

This paper focuses on end-to-end task-oriented dialogue systems,which jointly handle dialogue state tracking(DST)and response *** methods usually adopt a supervised paradigm to learn DST from a manually labeled ***,the annotation of the corpus is costly,time-consuming,and cannot cover a wide range of domains in the real *** solve this problem,we propose a multi-span prediction network(MSPN)that performs unsupervised DST for end-to-end task-oriented ***,MSPN contains a novel split-merge copy mechanism that captures long-term dependencies in dialogues to automatically extract multiple text spans as *** on these keywords,MSPN uses a semantic distance based clustering approach to obtain the values of each *** addition,we propose an ontology-based reinforcement learning approach,which employs the values of each slot to train MSPN to generate relevant *** results on single-domain and multi-domain task-oriented dialogue datasets show that MSPN achieves state-of-the-art performance with significant ***,we construct a new Chinese dialogue dataset MeDial in the low-resource medical domain,which further demonstrates the adaptability of MSPN.

关键词： end-to-end task-oriented dialogue dialogue state tracking(DST) unsupervised learning reinforcement learning

A V2V Emergent Message Dissemination Scheme for 6G-Oriented Vehicular Networks

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2023年第6期32卷 1179-1191页

作者： CHEN Chen WANG Chenyu CONG Li XIAO Ming PEI Qingqi School of Telecommunication Engineering Xidian University The State Grid Jilin Province Electric Power company limited Information Communication company KTH Royal Institute of Technology

To ensure traffic safety and improve traffic efficiency, vehicular networks come up with multiple types of messages for safety and efficiency applications. In sixth-generation(6G) systems, these messages should be timely and error-free disseminated through vehicle-to-vehicle(V2V) communication to ensure traffic safety and efficiency. V2V supports direct communication between two vehicle user equipments, regardless of whether a base station is involved. We propose a packet delivery ratio(PDR)-based message dissemination scheme(PDR-MD) between V2V in 6G-oriented vehicular networks to select relay vehicles when broadcasting emergent messages. This scheme grasps the balance between vehicle distance and PDR so as to reduce transmission delay while ensuring reliable PDR. We compared the PDR-MD scheme with other probabilistic broadcasting schemes. The experimental results show that the PDRMD protocol can maintain close to 95% and above PDR in transmitting emergent messages, and the transfer rate stays below 40%.

关键词： 6G mobile communication Protocols Urban areas Vehicular ad hoc networks Broadcasting Probabilistic logic Safety

An Artificial intelligence Algorithm for the Real-Time Early Detection of Sticking Phenomena in Horizontal Shale Gas Wells

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Fluid Dynamics & Materials Processing 2023年第10期19卷 2569-2578页

作者： Qing Wang Haige Wang Hongchun Huang Lubin Zhuo Guodong Ji CNPC Engineering Technology R&D company limited Planning and Support InstituteBeijing102206China

Sticking is the most serious cause of failure in complex drilling *** the present work a novel“early warning”method based on an artiﬁcial intelligence algorithm is proposed to overcome some of the known pro-blems associated with existing sticking-identiﬁcation *** method is tested against a practical case study(Southern Sichuan shale gas drilling operations).It is shown that the twelve sets of sticking fault diagnostic results obtained from a simulation are all consistent with the actual downhole state;furthermore,the results from four groups of veriﬁcation samples are also consistent with the actual downhole *** shows that the pro-posed training-based model can effectively be applied to practical situations.

关键词： Shale gas drilling sticking fault artificial intelligence risk early warning technology

Bottom hole pressure prediction based on hybrid neural networks and Bayesian optimization

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Petroleum Science 2023年第6期20卷 3712-3722页

作者： Chengkai Zhang Rui Zhang Zhaopeng Zhu Xianzhi Song Yinao Su Gensheng Li Liang Han College of Petroleum Engineering China University of Petroleum(Beijing)Beijing102249China College of Artificial intelligence China University of Petroleum(Beijing)Beijing102249China National Key Laboratory of Petroleum Resources and Engineering China University of Petroleum(Beijing)Beijing102249China CNPC Engineering Technology R&D company limited Beijing102206China

Many scholars have focused on applying machine learning models in bottom hole pressure (BHP) prediction. However, the complex and uncertain conditions in deep wells make it difficult to capture spatial and temporal correlations of measurement while drilling (MWD) data with traditional intelligent models. In this work, we develop a novel hybrid neural network, which integrates the Convolution Neural Network (CNN) and the Gate Recurrent Unit (GRU) for predicting BHP fluctuations more accurately. The CNN structure is used to analyze spatial local dependency patterns and the GRU structure is used to discover depth variation trends of MWD data. To further improve the prediction accuracy, we explore two types of GRU-based structure: skip-GRU and attention-GRU, which can capture more long-term potential periodic correlation in drilling data. Then, the different model structures tuned by the Bayesian optimization (BO) algorithm are compared and analyzed. Results indicate that the hybrid models can extract spatial-temporal information of data effectively and predict more accurately than random forests, extreme gradient boosting, back propagation neural network, CNN and GRU. The CNN-attention-GRU model with BO algorithm shows great superiority in prediction accuracy and robustness due to the hybrid network structure and attention mechanism, having the lowest mean absolute percentage error of 0.025%. This study provides a reference for solving the problem of extracting spatial and temporal characteristics and guidance for managed pressure drilling in complex formations.

关键词： Bottom hole pressure Spatial-temporal information Improved GRU Hybrid neural networks Bayesian optimization

维普期刊数据库