检索结果-内蒙古大学图书馆

36th Conference on neural Information processing Systems (NeurIPS)

作者： Sefidgaran, Milad Chor, Romain Zaidi, Abdellatif Huawei Technol France Paris Res Ctr Paris France Univ Paris Est F-77454 Champs Sur Marne France

ISBN: (纸本)9781713871088

In this paper, we use tools from rate-distortion theory to establish new upper bounds on the generalization error of statistical distributed learning algorithms. Specifically, there are K clients whose individually chosen models are aggregated by a central server. The bounds depend on the compressibility of each client's algorithm while keeping other clients' algorithms un-compressed, and leveraging the fact that small changes in each local model change the aggregated model by a factor of only 1/K. Adopting a recently proposed approach by Sefidgaran et al., and extending it suitably to the distributed setting, enables smaller rate-distortion terms which are shown to translate into tighter generalization bounds. The bounds are then applied to the distributed support vector machines (SVM), suggesting that the generalization error of the distributed setting decays faster than that of the centralized one with a factor of O(root log(K)/K). This finding is validated also experimentally. A similar conclusion is obtained for a multiple-round federated learning setup where each client uses stochastic gradient Langevin dynamics (SGLD).

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Research on Abnormal Power Consumption Behavior Detection and Load Forecasting System of Power Big Data Based on FCM and Wavelet neural network 14th

Research on Abnormal Power Consumption Behavior Detection an...

引用

14th International Conference on Frontier Computing, FC 2024

作者： Pan, Jingni Chai, Yufeng Lv, Jiayu Wang, Xiaotian State Grid Shizuishan Power Supply Co. Ningxia China State Grid Ningxia Power Supply Co. Ningxia China State Grid Wuzhong Power Supply Co. Ningxia China

ISBN: (纸本)9789819627974

With the progress of smart grid, power systems are faced with more and more complex data challenges, including large-scale data, multi-sample types and low resource integration. This research takes power big data as the object, carries out abnormal power consumption behavior detection and load prediction based on Hadoop platform, and successfully realizes the visualization processing of power data. We thoroughly analyzed the needs of power big data visualization and designed the overall Hadoop-based architecture. In terms of key technologies, we adopt the visualization method based on geographic information, and use MapReduce to carry out parallel computation, and effectively identify and display different electrical behaviors. In addition, we proposed a wavelet neural network power load forecasting method based on FCM, and personalized classification of power users through FCM clustering. Combined with the prediction ability of wavelet neural network, accurate power load prediction was successfully realized, and parallel calculation was carried out on the distributed platform to further improve the prediction accuracy. Finally, we built a Hadoop big data experiment platform to verify the experimental results and performance of the proposed method. The experimental results show that our research has the ability to effectively cope with the big data challenges of power system, and provides important support and guidance for the safe and stable operation of power system. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Abnormal Power Consumption Behavior Detection FCM Hadoop Platform Power Big Data Wavelet neural network

来源：评论

学校读者我要写书评

暂无评论

A distributed Real-time 3D Pose Estimation Framework based on Asynchronous Multiviews br

引用

KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS 2023年第2期17卷 559-575页

作者： Hwang, Taemin Kim, Jieun Kim, Minjoon Korea Elect Technol Inst Dept Data Convergence Platform Res Ctr Seongnam South Korea

3D human pose estimation is widely applied in various fields, including action recognition, sports analysis, and human-computer interaction. 3D human pose estimation has achieved significant progress with the introduction of convolutional neural network (CNN). Recently,several researches have proposed the use of multiview approaches to avoid occlusions in single-view approaches. However, as the number of cameras increases, a 3D pose estimation system relying on a CNN may lack in computational resources. In addition, when a single host system uses multiple cameras, the data transition speed becomes inadequate owing to bandwidth limitations. To address this problem, we propose a distributed real-time 3D pose estimation framework based on asynchronous multiple cameras. The proposed frameworkcomprises a central server and multiple edge devices. Each multiple-edge device estimates a 2D human pose from its view and sends it to the central server. Subsequently, the central server synchronizes the received 2D human pose data based on the timestamps. Finally, the central server reconstructs a 3D human pose using geometrical triangulation. We demonstrate that the proposed framework increases the percentage of detected joints and successfully estimates 3D human poses in real-time.

关键词： Computer vision edge processing multiple view geometry object detection pose estimation

来源：评论

学校读者我要写书评

暂无评论

An Effective Hybrid Model for Fake News Detection in Social Media Using Deep Learning Approach

引用

SN Computer Science 2024年第4期5卷 346页

作者： Raghavendra, R. Niranjanamurthy, M. Department of MCA BMS Institute of Technology and Management (Affiliated to Visvesvaraya Technological University) Jnana Sangama Avalahalli Karnataka Belgavi Bangalore 560064 India Department of AI and ML BMS Institute of Technology and Management (Affiliated to Visvesvaraya Technological University) Jnana Sangama Avalahalli Karnataka Belgavi Bangalore 560064 India

Social media is a strong Internet platform that allows people to voice their thoughts about numerous events happening in real time at multiple locations. People comment and express their thoughts on any social media post. Meanwhile, fake news or misleading information is disseminated regarding exciting occurrences that occur in real time. A large number of Internet users read and distribute such false and fake material without knowing the nature or legitimacy of the news. This has a detrimental influence on people’s perceptions of the particular event. Classical techniques are utilized to determine the type of news distributed on Twitter, Instagram, YouTube, and Facebook. However, these techniques fail to take into account characteristics such as news-generating location, consistency, timing, and novelty. It eventually leads to a scenario in which individuals form incorrect ideas and have misleading perceptions regarding any startling news. The spread of misleading thoughts and remarks has a significant impact on real-world action results. This research article addresses these difficulties by developing a system for detecting fake news using deep learning techniques and models. To determine the originality of the news, place of generation, and longevity, two deep learning models are developed artificial neural network and a mixed classifier model of convolution neural networks and long short-term memory. This aids in detecting bogus news and removing it from the server where it is stored. The tests utilizing these coupled models increase the detection of false information in social media. Furthermore, geo-map is used in this research to aids in the regulation of fake news flowing on social media regarding unique occasions occurring all over the world. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd 2024.

关键词： Artificial neural network (ANN) Convolutional neural network (CNN) False/fake news detection (FND) Long short-term memory (LSTM) Natural language processing (NLP) Stemming

来源：评论

学校读者我要写书评

暂无评论

Highly sensitive HF detection based on absorption enhanced light-induced thermoelastic spectroscopy with a quartz tuning fork of receive and shallow neural network fitting

引用

PHOTOACOUSTICS 2022年 28卷 100422页

作者： Liu, Xiaonan Qiao, Shunda Han, Guowei Liang, Jinxing Ma, Yufei Harbin Inst Technol Natl Key Lab Sci & Technol Tunable Laser Harbin 150001 Peoples R China Chinese Acad Sci Inst Semicond Engn Res Ctr Semicond Integrated Technol Beijing 100083 Peoples R China Southeast Univ Sch Instrument Sci & Engn Key Lab Microinertial Instrument & Adv Nav Technol Minist Educ Nanjing 210096 Peoples R China

Due to its advantages of non-contact measurement and high sensitivity, light-induced thermoelastic spectroscopy (LITES) is one of the most promising methods for corrosive gas detection. In this manuscript, a highly sensitive hydrogen fluoride (HF) sensor based on LITES technique is reported for the first time. With simple structure and strong robustness, a shallow neural network (SNN) fitting algorithm is introduced into the field of spectroscopy data processing to achieve denoising. This algorithm provides an end-to-end approach that takes in the raw input data without any pre-processing and extracts features automatically. A continuous wave (CW) distributed feedback diode (DFB) laser with an emission wavelength of 1.27 mu m was used as the excitation source. A Herriott multi-pass cell (MPC) with an optical length of 10.1 m was selected to enhance the laser absorption. A quartz tuning fork (QTF) with resonance frequency of 32,767.52 Hz was adopted as the thermoelastic detector. An Allan variance analysis was performed to demonstrate the system stability. When the integration time was 110 s, the minimum detection limit (MDL) was found to be 71 ppb. After the SNN fitting algorithm was used, the signal-to-noise ratio (SNR) of the HF-LITES sensor was improved by a factor of 2.0, which verified the effectiveness of this fitting algorithm for spectroscopy data processing.

关键词： Gas sensing Hydrogen fluoride Light-induced thermoelastic spectroscopy Shallow neural network Multi-pass cell

来源：评论

学校读者我要写书评

暂无评论

Satellite Image Classification using Convolutional neural network with Huber Loss Function for Land Cover Mapping in Forest Ecosystem

Satellite Image Classification using Convolutional Neural Ne...

引用

2024 International Conference on distributed Systems, Computer networks and Cybersecurity, ICDSCNC 2024

作者： Al-Fatlawy, Ramy Riad Raviprakash, P. Vybhavi, Gummagatta Yajaman Jagadish, Sripelli Sundaram, S Meenakshi College of Technical Engineering The Islamic University Department of Computers Techniques Engineering Najaf Iraq Hyderabad India Gokaraju Rangaraju Institute of Engineering and Technology Bachupally Department of Cse Hyderabad India School of Computer Science & Artificial Intelligence Sr University Warangal India Nitte Meenakshi Institute of Technology Department of Cse Bengaluru India

ISBN: (纸本)9798350375442

Satellite image classification for land cover involves determining high-resolution imagery to recognize and classify various types of land covers. This process assists in monitoring forest health, managing resources, and assessing biodiversity effectively. However, accurately classifying images for land cover is difficult because of diverse and complex nature of forest environments and varying image qualities. In this research, the Convolutional neural network with Huber Loss Function (CNN-HLF) is proposed to classify the satellite images accurately for landcover mapping in forest ecosystem. The CNN-HLF effectively manage the variability and complexity of forest environments which improves the performance of land cover. Initially, the Eurosat dataset is used to determine the CNN-HLF performance. Resizing and image sharpening is applied to adjust the size and enhance the clarity of features. DenseNet201 is employed for extracting features effectively from pre-processed images. The proposed CNN-HLF achieves a high accuracy of 99.87% using Eurosat dataset compared to existing methods like GoogleNet with pre-processing and Lightweight Parallel CNN (LPCNN). © 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

Computationally Efficient neural Rendering for Generator Adversarial networks Using a Multi-GPU Cluster in a Cloud Environment

引用

IEEE ACCESS 2023年 11卷 45559-45571页

作者： Ravikumar, Aswathy Sriraman, Harini Vellore Inst Technol Sch Comp Sci & Engn Chennai 600127 India

Due to its fantastic performance in the quality of the images created, Generator Adversarial networks have recently become a viable option for image reconstruction. The main problem with employing GAN is how expensive the computations are. Researchers have developed techniques for distributing GANs across multiple nodes. However, these techniques typically do not scale because they frequently separate the components (Discriminator and Generator), leading to high communication overhead or encountering distribution-related problems unique to GAN training. In this study, the training procedure for the GAN is parallelized and carried out over many Graphical processing Units (GPUs). TensorFlow's built-in logic and a custom loop were tweaked for more control over the resources allotted to each GPU worker. In this study, GPU image processing improvements and multi-GPU learning are used. The GAN model is accelerated using distributed TensorFlow with synchronous data-parallel training on a single system and several GPUs. Acceleration was accomplished using the Genesis Cloud Platform and the NVIDIA CIRCLED LATIN CAPITAL LETTER R GeForceTM GTX 108 GPU accelerator. The speed-up of 1.322 for two GPUs, 1.688 for three GPUs, and 1.7792 for four GPUs using multi-GPU acceleration. The parameter server model's data initialization and image production bottlenecks are removed, but the results' speed-up is not linear. Increasing the number of GPUs and removing the connectivity constraint will accelerate things even more. The bottlenecks are detected using new network lines and resources, and solutions are suggested. Recomputation and quantization are the two techniques to reduce the amount of GPU acceleration in memory. Deployment and versioning are essential for successfully operating multi-node GAN models in MLflow. Properly deploying and versioning these models can improve scalability, reproducibility, and collaboration across teams working on the same model. MLflow provides built-in

关键词： Training Computational modeling Generative adversarial networks Graphics processing units Data models Load modeling Servers All reduce bottleneck data parallel fault tolerance generative adversarial network GPU parallel learning cloud computing

来源：评论

学校读者我要写书评

暂无评论

Impact of multi-GPUs and hyperparameters on parallel algorithms based on pytorch distributed 3

Impact of multi-GPUs and hyperparameters on parallel algorit...

引用

3rd International Conference on Signal processing and Machine Learning, CONF-SPML 2023

作者： Wei, Shiyu SWUFE-UD Institute of Data Science Southwestern University of Finance and Economics Chengdu610000 China

neural networks, as powerful models for solving many genuine problems, are suffering from the issue of being too computationally intensive due to the excessive size of the models themselves and the large datasets used for training. This paper's work focuses on a parallel algorithm, experimented on the classical convolutional neural network LeNet-5, aiming to reduce the time it takes for the neural network to train. This applies a massively parallel GPU structure to take full advantage of its high computing power. Among the approaches to improve the performance of parallel algorithms, many efforts have been made to improve the design of parallel structures, while this work focuses more on the enhancement of parallel algorithms by reasonable settings of neural network hyperparameters. This paper tries to approach (1) reducing the running time by testing parallelism with different numbers of GPUs and (2) further investigating the relationship between parallelism and model accuracy degradation and experimentally deriving a parallel scheme that can reduce the training time while guaranteeing accuracy. © Published under licence by IOP Publishing Ltd.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Graph neural network Based on Spectral Parameter Estimation for Mobile networks 10

Graph Neural Network Based on Spectral Parameter Estimation ...

引用

10th International Conference on Engineering and Emerging Technologies, ICEET 2024

作者： Jin, Renhong Yan, Xin School of Computer Science and Artificial Intelligence Wuhan University of Technology Wuhan430070 China

ISBN: (纸本)9798331532895

For online learning, Graph neural networks(GNNs) often require real graph data for parameter fine-tuning, but obtaining the graph data is challenging. Currently, GNNs are trained based on existing graph data. When learning in dynamic networks, spatial-domain GNNs use the adjacency matrix of entire graphs as its input, which are extremely resource-intensive in terms of computation and storage while dealing with large-scale graph data. Moreover, spectral-domain GNNs suffer from high spectral decomposition complexity. To enhance the generalization of GNN models and achieve lightweight GNN models, this paper proposes a lightweight GNN model based on spectral domain, SDGAT(Spectral-Domain Graph Attention network). By combining spectral domain graph convolution and attention mechanism, SDGAT can effectively capture the global structure and local features of a graph, which helps to suppress noise in the graph and improve the robustness of the model. Furthermore, a distributed estimation algorithm for dynamic network spectral parameters is proposed, which significantly reduces the computational overhead when processing large-scale graph data. Finally, authors validate the effectiveness and accuracy of SDGAT through classification and regression tasks, and explore the feasibility of extending this model to other GNNs. © 2024 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

ResNet110 and Mask Recurrent Convolutional neural network Based Detection and Classification of Colorectal Cancer 2

ResNet110 and Mask Recurrent Convolutional Neural Network Ba...

引用

2nd International Conference on distributed Computing and Optimization Techniques, ICDCOT 2024

作者： Alzubaidi, Laith H. Priyanka, K. Suhas, S.K. Kumar, B. Kiran Malathy, V. The Islamic University Najaf Iraq Nitte Meenakshi Institute of Technology Dept of ISE Bengaluru India S.E.A College of Engineering and Technology Department of Electronics and Communication Engineering Bangalore India Cvr College of Engineering Department of Information Technology Telangana Hyderabad India Sr University Department of Ece Warangal India

ISBN: (纸本)9798350382952

Globally, Colorectal Cancer (CRC) is the one most significant cancer types as well as it grows in a region of colon of large intestine. An early CRC detection is supportive to handle the impression of accumulating cancer cells. However, the detection process of utilizing the colon images is expensive as well as time-consuming. Recently, Deep Learning (DL)-based approaches are developed for the CRC detection. Therefore, this research proposes the efficient pre-trained architecture for the detection and classification of CRC. Initially, the benchmark dataset of Warwick-QU is collected for the estimation of the effectiveness of the proposed method. Then, the pre-processing is done by the utilization of Noise removal as well as Contrast Enhancement. Then, pre-processed data is to extracted by the utilization the pre-trained architecture. Finally, the outcome of extracted features is classified by the utilization of improved Mask Recurrent-Convolutional neural network (Mask R-CNN). The effectiveness of the proposed method is validated by various performance metrices and it achieves the accuracy of 99.67%, specificity of 99.56%, sensitivity of 99.57%, precision of 97.36% and F1-score of 98.34% when compared to the existing methods such as ResNet-50, Stacked Gated Recurrent Unit (SGRU) and Mobile Net. © 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

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：