检索结果-内蒙古大学图书馆

NOISE SUPPRESSION FOR DAS RECORDS USING THE DOUBLE-SCALE DOUBLE-PATH ATTENTION network

INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL 2024年第3期20卷 911-928页

作者： Zhong, Tie Yu, Hang Wang, Wei Dong, Shiqi Minist Educ Key Lab Modern Power Syst Simulat & Control & Rene 169Changchun Rd Jilin 132012 Peoples R China Northeast Elect Power Univ Sch Elect Engn 169Changchun Rd Jilin 132012 Peoples R China Northwest Branch China Petr Explorat Dev Res Inst Inst Geophys 561Yanerwan Rd Lanzhou 730030 Peoples R China

. distributed acoustic sensing (DAS) is an innovative acquisition technology that has gradually found application in seismic exploration due to its superior acquisition capabilities. However, the significant amplitude disparity among direct waves, reflected waves, and converted waves in DAS data, coupled with the presence of strong noise pollution, poses major challenges to the subsequent analysis and interpretation of DAS data, ultimately impeding the application and development of the technology. Addressing the need for precise denoising of DAS data, we propose a double-scale double-path attention network (DSDPAN), which extracts effective features through the collaborative representation of information across different scales to enhance its feature expression ability. Under low-scale conditions, DSDPAN employs a method that constructs multiple denoising paths and fuses them based on their correlation to achieve complementary advantages, thereby enhancing denoising performance. Moreover, the spatial attention mechanism is introduced to redistribute the weight of collected features, prioritizing those that are more conducive to noise reduction. The experimental results confirm that DSDPAN outperforms other methods in processing DAS data. After training on a DAS data set consisting noise suppression capabilities but also achieves effective signal recovery.

关键词： distributed acoustic sensing (DAS) Background noise suppression Convolutional neural network (CNN)

来源：评论

学校读者我要写书评

暂无评论

Sports Event Detection Using a 3D Convolutional neural network with Equiangular Basis Vector in Video processing 3

Sports Event Detection Using a 3D Convolutional Neural Netwo...

引用

3rd IEEE International Conference on distributed Computing and Electrical Circuits and Electronics, ICDCECE 2024

作者： Kumar, M. Ashok Jasim, Laith H. Sudhakar, K. Reddy, R. Archana Saranya, N. Naga Skyline University Nigeria Department Of Computer Science Software Engineering And Cyber Security Kano State Nigeria College Of Technical Engineering The Islamic University Department Of Computers Techniques Engineering Najaf Iraq Nitte Meenakshi Institute Of Technology Dept. Of Ai & Ds Bengaluru India School Of Computer Science And Engineering Sr University Warangal India Saveetha College Of Liberal Arts & Science Simats Department Of Computer Applications Chennai India

ISBN: (纸本)9798350318609

As sports video collections continue to expand at a rapid pace, there is a need for advanced classification methods that can accurately and efficiently categorize and manage the massive amount of content available. These methods must be able to dynamically adapt to changes in the collection, ensuring that the videos are properly organized and easily searchable for users. Traditional classification methods are not able to keep up with the rapid growth and complexity of sports video collections, leading to incomplete classification and impacting search and retrieval processes. This paper addresses this challenge using the UCF sports dataset and proposes a Visual Geometry Group-19 (VGG-19) based CNN classifier and Equiangular Basis Vector (EBV) representation introduced for efficient classification and category representation. Furthermore, a 3D Convolutional neural network (3D-CNN) architecture extracts both spatial and temporal features from videos for accurate action classification. The Adam optimizer in the optimization algorithm is employed to optimize the learning process. This paper shows that the proposed 3D-CNN model outperforms existing methods a Hybrid Convolutional neural network with a Bidirectional Recurrent neural network (HCNN-BiRNN) model, Deformable Convolution networks (DCN), 2D-Convolutional neural network (2D-CNN), Convolutional neural network - Long Short-Term Frequency (CNN-LSTM) in terms of accuracy (99.45%), precision (98.00%), sensitivity (96.98%), and F1-score (97.05%). © 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

An empirical study on the structure evolution of deep learning models: taking SAR image processing a case study 14

An empirical study on the structure evolution of deep learni...

引用

IEEE 14th International Conference on Joint Cloud Computing (JCC)

作者： Liu, Huanxi He, Xiang Feng, Dawei Bao, Han Natl Univ Def Technol Natl Lab Parallel & Distributed Proc PDL Changsha Peoples R China Natl Univ Def Technol Sch Informat & Commun Changsha Peoples R China

ISBN: (纸本)9798350328554

With the continuous improvement on model performance, deep learning models have been widely deployed and achieved promising outcomes in various fields in recent years. However, due to the escalating volumes of training data and the complexity of application problems, it becomes more and more challenging to design a neural network with better performance by hand. Analysing the evolution of typical neural network structures has important reference significance for designing a network structure. In this paper, we select the open source models in SAR image processing for an empirical analysis on the evolution of neural network structures. We analyse the evolution of 239 open source deep learning models from the aspects of framework, computing unit, model computation amount and the combined use of various computing units. Results reveal that preference and co-occurrence exist in computing units, while the average number of convolution, activation and normalization layer increases significantly over time. Model complexity shows an overall upward trend, and the characteristics of SAR image are more and more taken into consideration during the model structure design.

关键词： Intelligent model evolution neural network structure analysis SAR image processing

来源：评论

学校读者我要写书评

暂无评论

WholeGraph: A Fast Graph neural network Training Framework with Multi-GPU distributed Shared Memory Architecture

WholeGraph: A Fast Graph Neural Network Training Framework w...

引用

Supercomputing Conference

作者： Dongxu Yang Junhong Liu Jiaxing Qi Junjie Lai NVIDIA Beijing China

ISBN: (纸本)9781665454452

Graph neural networks (GNNs) are prevalent to deal with graph-structured datasets, encoding graph data into low dimensional vectors. In this paper, we present a fast training graph neural network framework, i.e., WholeGraph, based on a multi-GPU distributed shared memory architecture. Whole-Graph consists of partitioning the graph and corresponding node or edge features to multi-GPUs, eliminating the bottleneck of communication between CPU and GPUs during the training process. And the communication between different GPUs is implemented by GPUDirect Peer-to-Peer (P2P) memory access technology. Furthermore, WholeGraph provides several optimized computing operators. Our evaluations show that on large-scale graphs WholeGraph outperforms state-of-the-art GNN frameworks, such as Deep Graph Library (DGL) and Pytorch Geometric (PyG). The speedups of WholeGraph are up to $57.32\mathrm{x}$ and $242.98\mathrm{x}$ compared with DGL and PyG on a single machine multi-GPU node, respectively. The ratio of GPU utilization can sustain above 95% during GNN training process.

关键词： Training Computational modeling High performance computing Memory architecture Graphics processing units distributed databases Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Secure Artificial Intelligence for Precise Vehicle Behavior Prediction in 6G Consumer Electronics

引用

IEEE TRANSACTIONS ON CONSUMER ELECTRONICS 2024年第1期70卷 3898-3905页

作者： Haider, Sami Ahmed Ramesh, Janjhyam Venkata Naga Raina, Vikas Maaliw III, Renato R. Soni, Mukesh Nasurova, Kamolakhon Patni, Jagdish Chandra Singh, Pavitar Parkash Univ Glasgow Glasgow Coll Glasgow G12 8QQ Scotland Koneru Lakshmaiah Educ Fdn Dept Comp Sci & Engn Vaddeswaram 522502 India Mody Univ Sci & Technol CSE SET Sikar 332311 India Southern Luzon State Univ Coll Engn Lucban 4328 Philippines Dr DY Patil Sch Sci & Technol Pune India Tashkent Inst Finance Dept Sci Res Innovat & Training Sci & Pedag Person Tashkent 1000000 Uzbekistan Alliance Univ Bengaluru Sch Adv Comp CSE Anekal 562106 India Lovely Profess Univ Dept Management Phagwara 144001 India

In the context of Secure Artificial Intelligence for 6G Consumer Electronics, accurately predicting vehicle behavior in dynamic traffic scenarios is a significant challenge in intelligent transportation. To avoid sending all raw data to a centralized cloud server, this study presents an artificial intelligence (AI) based distributed machine learning framework (AICEML) that can run on local edge devices. This method protects user privacy while minimizing transmission and processing delays. Accurate predictions are maintained despite the presence of many cars because to AICEML's use of the model on edge devices, which incorporates edge-enhanced attention and graph convolutional neural network features to swiftly collect and transmit vehicle interaction information. Each edge device can adapt its neural network type and scale based on its computing capabilities, accommodating various application scenarios. Experimental results using the NGGSIM dataset demonstrate AICEML's superiority, achieving precision, recall, and F1 scores of 0.9391, 0.9557, and 0.9473, respectively. With a 1-second prediction horizon, it maintains 91.21% accuracy and low time complexity even as the number of vehicles increases. This framework holds promise for enhancing intelligent transportation systems in the 6G era while prioritizing security and efficiency.

关键词： Behavioral sciences Servers Data models Training Computational modeling Hidden Markov models neural networks Secure artificial intelligence 6G consumer electronics vehicle behavior prediction edge computing communication efficiency

来源：评论

学校读者我要写书评

暂无评论

Data Valuation Methods for Federated Learning 31

Data Valuation Methods for Federated Learning

引用

31st IEEE Conference on Signal processing and Communications Applications (SIU)

作者： Ardic, Emre Genc, Yakup Gebze Tekn Univ Bilgisayar Muhendisligi Bolumu Kocaeli Turkiye

ISBN: (纸本)9798350343557

Modern distributed networks such as smartphones, wearable devices, and self-driving vehicles generate a wealth of data. As the computation, storage, and battery capabilities of these devices grow, local data storage and processing become easier and more secure. This has led to a growing interest in federated learning which provides training of deep learning models while keeping the training data decentralized. During the training, the performance of deep learning models can be improved by filtering redundant, malicious, and abnormal samples with data valuation methods. In this work, the aim is to improve federated learning models by integrating data valuation methods into the training process. For this, a two-layered convolutional neural network is trained on the MNIST dataset inside a small federated learning network. The Shapley and Leave-one-out based data evaluation methods developed in this study have resulted in an 8.81% accuracy improvement during experiments conducted on the MNIST dataset with corrupted labels.

关键词： Federated Learning (FL) Data Valuation Convolutional neural networks (CNN)

来源：评论

学校读者我要写书评

暂无评论

Prediction of Depression Severity Based on the Prosodic and Semantic Features With Bidirectional LSTM and Time distributed CNN

引用

IEEE TRANSACTIONS ON AFFECTIVE COMPUTING 2023年第3期14卷 2251-2265页

作者： Mao, Kaining Zhang, Wei Wang, Deborah Baofeng Li, Ang Jiao, Rongqi Zhu, Yanhui Wu, Bin Zheng, Tiansheng Qian, Lei Lyu, Wei Ye, Minjie Chen, Jie Univ Alberta Dept Elect & Comp Engn Edmonton AB T6G 2R3 Canada Wenzhou Med Univ Affiliated Kangning Hosp Wenzhou 325015 Zhejiang Peoples R China Wenzhou Med Univ Wenzhou 325015 Zhejiang Peoples R China

Depression is increasingly impacting individuals both physically and psychologically worldwide. It has become a global major public health problem and attracts attention from various research fields. Traditionally, the diagnosis of depression is formulated through semi-structured interviews and supplementary questionnaires, which makes the diagnosis heavily relying on physicians' experience and is subject to bias. However, since the pathogenic mechanism of depression is still under investigation, it is difficult for physicians to diagnose and treat, especially in the early clinical stage. As smart devices and artificial intelligence advance rapidly, understanding how depression associates with daily behaviors can be beneficial for the early stage depression diagnosis, which reduces labor costs and the likelihood of clinical mistakes as well as physicians bias. Furthermore, mental health monitoring and cloud based remote diagnosis can be implemented through an automated depression diagnosis system. In this article, we propose an attention-based multimodality speech and text representation for depression prediction. Our model is trained to estimate the depression severity of participants using the Distress Analysis Interview Corpus-Wizard of Oz (DAIC-WOZ) dataset. For the audio modality, we use the collaborative voice analysis repository (COVAREP) features provided by the dataset and employ a Bidirectional Long Short-Term Memory network (Bi-LSTM) followed by a Time-distributed Convolutional neural network (T-CNN). For the text modality, we use global vectors for word representation (GloVe) to perform word embeddings and the embeddings are fed into the BiLSTM network. Results show that both audio and text models perform well on the depression severity estimation task, with best sequence level F-1 score of 0.9870 and patient-level F-1 score of 0.9074 for the audio model over five classes (healthy, mild, moderate, moderately severe, and severe), as well as sequence level

关键词： Artificial intelligence depression machine learning mental health natural language processing neural network

来源：评论

学校读者我要写书评

暂无评论

Machine Learning-empowered network Measurement: A Critical Path to Traffic Anomaly Detection in IoT-enabled Smart Grid 30

Machine Learning-empowered Network Measurement: A Critical P...

引用

30th IEEE International Conference on Parallel and distributed Systems, ICPADS 2024

作者： Zang, Hongrui Liu, Hongbo Sun, Wei Zhang, Guangyuan Kong, Xiangyu State Grid Jilin Electric Power Company Limited Communication Company Branch Changchun China Dalian University of Technology School of Computer Science and Technology Dalian China

ISBN: (纸本)9798331515966

Large-scale deployment of Internet of Things (IoT) devices provides efficient data collection and control capabilities in the smart grid, while edge computing plays a key role in increasing the speed of data processing and reducing latency. The growth of edge computing in the smart grid is inevitably increasing the stability requirement in network transmission and the necessity of adequate network measurement to detect traffic anomalies. There are two challenges for measurement in IoT-enabled networks. The first challenge is the high processing speed and limited memory space. The second challenge is the varying network traffic. This paper studies how to use sketch techniques to fulfill the required capabilities of network measurement. Sketches have been considered as the most promising solution for network measurement in recent years, because they greatly optimize the speed and memory usage at the cost of small error. However, most sketches do not work well for varying network traffic. These sketches require to adjust their internal memory usage while the optimal point is sensitive to specified flow size distribution and memory size. To address this problem, we propose a machine learning empowered sketch framework. The proposed sketch framework trains a neural network model that online adjusts the memory usage based on a small sample of flow. We conduct data-driven simulations to evaluate the proposed framework. By comparing measurement results on public dataset, we can see that the proposed framework significantly improves the accuracy of measurement and anomaly detection without sacrificing the line-rate processing capability. © 2024 IEEE.

关键词： Edge computing

来源：评论

学校读者我要写书评

暂无评论

Exponential Quantum Communication Advantage in distributed Inference and Learning 38

Exponential Quantum Communication Advantage in Distributed I...

引用

38th Conference on neural Information processing Systems, NeurIPS 2024

作者： Gilboa, Dar Michaeli, Hagay Soudry, Daniel McClean, Jarrod R. Google Quantum AI VeniceCA United States Technion Haifa Israel

Training and inference with large machine learning models that far exceed the memory capacity of individual devices necessitates the design of distributed architectures, forcing one to contend with communication constraints. We present a framework for distributed computation over a quantum network in which data is encoded into specialized quantum states. We prove that for models within this framework, inference and training using gradient descent can be performed with exponentially less communication compared to their classical analogs, and with relatively modest overhead relative to standard gradient-based methods. We show that certain graph neural networks are particularly amenable to implementation within this framework, and moreover present empirical evidence that they perform well on standard benchmarks. To our knowledge, this is the first example of exponential quantum advantage for a generic class of machine learning problems that hold regardless of the data encoding cost. Moreover, we show that models in this class can encode highly nonlinear features of their inputs, and their expressivity increases exponentially with model depth. We also delineate the space of models for which exponential communication advantages hold by showing that they cannot hold for linear classification. Communication of quantum states that potentially limit the amount of information that can be extracted from them about the data and model parameters may also lead to improved privacy guarantees for distributed computation. Taken as a whole, these findings form a promising foundation for distributed machine learning over quantum networks. © 2024 neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A novel multi-modal Federated Learning based thermal-aware job scheduling framework

引用

COMPUTER networkS 2024年 255卷

作者： Rehman, Rameesha Malik, Saif Ur Rehman Qureshi, Shahida Hafeezan Atif Mouqarrab, Syed COMSATS Univ Islamabad Dept CS Islamabad Pakistan Univ Dublin Trinity Coll Sch Comp Sci & Stat Dublin Ireland Univ Bedfordshire Sch Comp Sci & Technol Luton LU1 3JU England

Cooling costs constitute more than half of the total data center energy expenditure. Thermal imbalance results in hotspot regions requiring additional cooling power. To reduce it, thermal aware job scheduling is a wellknown software solution that is subject to predicting correct server temperatures. Existing solutions have not explored intelligent solutions and rely only on logic based algorithms to allocate tasks that work on predefined rules. Few deep learning based solutions that are proposed, have not explored its alternatives and existing data modalities in data centers, resulting in inefficient models. Existing literature only proposes solutions based on unimodal tabular data. Therefore, we propose a multimodal architecture that considers different underlying data modalities in data centers to increase the model's efficiency and predict correct server temperatures. The increasing production of data and the need for storage and processing units has led to the development of distributed data centers. Existing techniques are limited to individual data centers which fail to consider the data privacy restrictions that arise while dealing with distributed scenarios. Findings from our simulations affirm our proposed scheme in terms of the objectives mentioned above. We propose a federated learning architecture that efficiently deals with distributed data centers while ensuring privacy. Our simulation results show an overall increase in the efficiency of the model in comparison to an existing intelligent solution. Furthermore, we provide comparative results that show how our model performs better and achieves lower thermal imbalance as compared to an existing scheme.

关键词： distributed data centers Thermal-aware job scheduling Data center privacy Temperature prediction Deep neural network Federated Learning Heatmap Multimodal Thermal imbalance

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：