检索结果-南通市图书馆

Loss Aware Feature Attention Mechanism for Class and Feature Imbalance Issue

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2025年第1期82卷 751-775页

作者： Yuewei Wu Ruiling Fu Tongtong Xing Fulian Yin College of information and communication Engineering Communication University of ChinaBeijing100024China State Key Laboratory of Media Convergence and communication Communication University of ChinaBeijing100024China

In the Internet era,recommendation systems play a crucial role in helping users find relevant information from large *** imbalance is known to severely affect data quality,and therefore reduce the performance of recommendation *** to the imbalance,machine learning algorithms tend to classify inputs into the positive(majority)class every time to achieve high prediction *** can be categorized such as by features and classes,but most studies consider only class *** this paper,we propose a recommendation system that can integrate multiple networks to adapt to a large number of imbalanced features and can deal with highly skewed and imbalanced datasets through a loss *** propose a loss aware feature attention mechanism(LAFAM)to solve the issue of feature *** network incorporates an attention mechanism and uses multiple sub-networks to classify and learn *** better results,the network can learn the weights of sub-networks and assign higher weights to important *** propose suppression loss to address class imbalance,which favors negative loss by penalizing positive loss,and pays more attention to sample points near the decision *** on two large-scale datasets verify that the performance of the proposed system is greatly improved compared to baseline methods.

关键词： Imbalanced data deep learning e-commerce recommendation loss function big data analysis

Software-defined process-near-memory architecture using 3D hybrid bonding integration

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(information Sciences) 2025年第1期68卷 366-378页

作者： Anlin XU Chenchen DENG Jianfeng ZHU Yao WANG Shaojun WEI Leibo LIU school of Integrated Circuits Tsinghua University Beijing Institution of Tracking and communication Technology Beijing National Research Center for information Science and Technology Tsinghua University

With the unprecedented explosive growing amount of global data, the development of computing chips, which encounter bottlenecks such as power wall and memory wall, cannot satisfy the demanding requirement. This work proposes a software-defined process-near-memory(SDPNM) computing architecture implemented using 3D hybrid bonding *** software-defined chip architecture, featuring spatial computations and dynamic reconfiguration, innovates in a top-down manner to achieve high energy efficiency while maintaining flexibility after fabrication. The process-near-memory integration further advances the SDNPM chip in a bottom-up way to reduce the energy consumption of data movement while improving the bandwidth. Utilizing a relatively mature fabrication and bonding process can result in feasible solutions for both data-intensive and compute-intensive applications including digital signal processing and artificial intelligence. The logic die is fabricated in the SMIC 40 nm process and the DRAM die is fabricated in the PSMC 25 nm process. The hybrid bonding is implemented by XMC. The experimental results show that the energy efficiency of the proposed SDPNM chip is 33.1× better than the state-of-the-art FPGA ranging from 8.2× to 104.1×.

关键词： software-defined chips process near memory energy efficiency dynamic reconfiguration domain specific

同方期刊数据库

Steel Surface Defect Recognition in Smart Manufacturing Using Deep Ensemble Transfer Learning-Based Techniques

在线全文

学校读者我要写书评

暂无评论

Computer Modeling in Engineering & Sciences 2025年第1期142卷 231-250页

作者： Tajmal Hussain Jongwon Seok Department of information and communication Engineering Changwon National UniversityChangwon51140Republic of Korea

Smart manufacturing and Industry 4.0 are transforming traditional manufacturing processes by utilizing innovative technologies such as the artificial intelligence(AI)and internet of things(IoT)to enhance efficiency,reduce costs,and ensure product *** light of the recent advancement of Industry 4.0,identifying defects has become important for ensuring the quality of products during the manufacturing *** this research,we present an ensemble methodology for accurately classifying hot rolled steel surface defects by combining the strengths of four pre-trained convolutional neural network(CNN)architectures:VGG16,VGG19,Xception,and Mobile-Net V2,compensating for their individual *** evaluated our methodology on the Xsteel surface defect dataset(XSDD),which comprises seven different *** ensemble methodology integrated the predictions of individual models through two methods:model averaging and weighted *** evaluation showed that the model averaging ensemble achieved an accuracy of 98.89%,a recall of 98.92%,a precision of 99.05%,and an F1-score of 98.97%,while the weighted averaging ensemble reached an accuracy of 99.72%,a recall of 99.74%,a precision of 99.67%,and an F1-score of 99.70%.The proposed weighted averaging ensemble model outperformed the model averaging method and the individual models in detecting defects in terms of accuracy,recall,precision,and *** analysis with recent studies also showed the superior performance of our methodology.

关键词： Smart manufacturing CNN steel defects ensemble models

Unmasking Social Robots’Camouflage:A GNN-Random Forest Framework for Enhanced Detection

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2025年第1期82卷 467-483页

作者： Weijian Fan Chunhua Wang Xiao Han Chichen Lin school of Data Science and Intelligent Media Communication University of ChinaBeijing100024China school of Computer and Cyber Sciences Communication University of ChinaBeijing100024China Institute of communication Studies Communication University of ChinaBeijing100024China State Key Laboratory of Media Convergence and communication Communication University of ChinaBeijing100024China

The proliferation of robot accounts on social media platforms has posed a significant negative impact,necessitating robust measures to counter network anomalies and safeguard content *** robot detection has emerged as a pivotal yet intricate task,aimed at mitigating the dissemination of misleading *** graphbased approaches have attained remarkable performance in this realm,they grapple with a fundamental limitation:the homogeneity assumption in graph convolution allows social robots to stealthily evade detection by mingling with genuine human *** unravel this challenge and thwart the camouflage tactics,this work proposed an innovative social robot detection framework based on enhanced HOmogeneity and Random Forest(HORFBot).At the core of HORFBot lies a homogeneous graph enhancement strategy,intricately woven with edge-removal techniques,tometiculously dissect the graph intomultiple revealing ***,leveraging the power of contrastive learning,the proposed methodology meticulously trains multiple graph convolutional networks,each honed to discern nuances within these tailored *** culminating stage involves the fusion of these feature-rich base classifiers,harmoniously aggregating their insights to produce a comprehensive detection *** experiments on three social robot detection datasets have shown that this method effectively improves the accuracy of social robot detection and outperforms comparative methods.

关键词： Social robot detection graph neural networks random forest homophily heterophily

Turbulent Agglomeration of Microparticles in a Cylinder Wake Flow Using LES-DEM:Focusing on the Effect of the Reynolds Number

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 Thermal Science 2025年第1期34卷 34-49页

作者： WANG Shuang MU Lin LI Xue XIE Jun DONG Ming school of Energy and Power Engineering Dalian University of Technology school of Optical information and Energy Engineering Wuhan Institute of Technology school of Energy and Environment Shenyang Aerospace University

Turbulent agglomeration is viewed as a promising technology for enhancing fine particle removal *** better understand particle transport,agglomeration behaviors,and fluid-particle interactions,we numerically explored these phenomena under cylindrical vortex wake influence using a coupled large eddy simulation and discrete element method（LES-DEM） *** validity of the LES approach was verified by comparison with available direct numerical simulation（DNS） *** adopted the Johnson-Kendall-Roberts（JKR） contact model for particle-particle *** particle dispersion and agglomeration characteristics of particles with different diameters（dp=2-20 μm） in the laminar and transition of shear layer（TrSL） flow regimes were *** particles were concentrated at the vortex centers,while larger particles accumulated around the *** agglomeration efficiency exhibited an M-shaped profile spanwise（y-direction）.With increasing Reynolds number,the agglomeration efficiency and turbulence intensity *** particle agglomeration efficiency peaks at a certain Reynolds ***,at higher Reynolds numbers,reducing the residence time of particles in the flow field decreases the agglomeration efficiency.

关键词： discrete element method large eddy simulation microparticle particle agglomeration

Multi-receptive field interaction network for shape from polarization

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(information Sciences) 2025年第1期68卷 394-395页

作者： Yini PENG Rui LIU Zhiyuan ZHANG Zhongyuan WANG Jiayi MA Xin TIAN Electronic information school Wuhan University Computer Science school Wuhan University Wuhan Institute of Quantum Technology

Shape from polarization (SfP) method can use the polarization information in reflected light to estimate the surface normal of the target,which can further reconstruct the shape of the *** a simple image capture process,it can use low-cost equipment to meet a high precision imaging requirement,which can be used in remote scenes and other applications.

关键词：

IDSSCNN-XgBoost:Improved Dual-Stream Shallow Convolutional Neural Network Based on Extreme Gradient Boosting Algorithm for Micro Expression Recognition

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2025年第1期82卷 729-749页

作者： Adnan Ahmad Zhao Li Irfan Tariq Zhengran He school of information Science and Engineering Southeast UniversityNanjing210016China

Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial *** deep neural networks(DNNs)with convolutional structures have been ***,unlike DNNs,shallow convolutional neural networks often outperform deeper models in mitigating overfitting,particularly with small ***,many of these methods rely on a single feature for recognition,resulting in an insufficient ability to extract highly effective *** address this limitation,in this paper,an Improved Dual-stream Shallow Convolutional Neural Network based on an Extreme Gradient Boosting Algorithm(IDSSCNN-XgBoost)is introduced for ME *** proposed method utilizes a dual-stream architecture where motion vectors(temporal features)are extracted using Optical Flow TV-L1 and amplify subtle changes(spatial features)via EulerianVideoMagnification(EVM).These features are processed by IDSSCNN,with an attention mechanism applied to refine the extracted effective *** outputs are then fused,concatenated,and classified using the XgBoost *** comprehensive approach significantly improves recognition accuracy by leveraging the strengths of both temporal and spatial information,supported by the robust classification power of *** proposed method is evaluated on three publicly available ME databases named Chinese Academy of Sciences Micro-expression Database(CASMEII),Spontaneous Micro-Expression Database(SMICHS),and Spontaneous Actions and Micro-Movements(SAMM).Experimental results indicate that the proposed model can achieve outstanding results compared to recent *** accuracy results are 79.01%,69.22%,and 68.99%on CASMEII,SMIC-HS,and SAMM,and the F1-score are 75.47%,68.91%,and 63.84%,*** proposed method has the advantage of operational efficiency and less computational time.

关键词： ME recognition dual stream shallow convolutional neural network euler video magnification TV-L1 XgBoost

Experimental demonstration of 220-GHz terahertz signals wireless transmission over 4.6 km

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(information Sciences) 2025年第1期68卷 386-388页

作者： Yi WEI Jianjun YU Mingxu WANG Xiongwei YANG Qiutong ZHANG Jingwen TAN Bing ZHANG Ying WU Yang HAN Peng TIAN Wen ZHOU Kaihui WANG Feng ZHAO State Key Laboratory of ASIC and System Key Laboratory for Information Science of Electromagnetic Waves (MoE)School of Information Science and Technology Fudan University school of Electronic Engineering Xi'an University of Posts and Telecommunications

The advancement of terahertz (THz) communication technology drives the evolution of wireless communication systems,offering novel pathways and technical means for the development of future 6G communication *** wireless communication systems are often constrained by bandwidth limitations of electronic devices in high frequency ***,THz communication technology leverages the characteristics of electromagnetic waves to transcend these limitations,enabling communication athigher frequencies and wider bandwidths.

关键词：

同方期刊数据库

Federated local causal structure learning

在线全文

学校读者我要写书评

暂无评论

Science China(information Sciences) 2025年

作者： Kui YU Chen RONG Hao WANG Fuyuan CAO Jiye LIANG school of Computer and information Hefei University of Technology school of Computer and information Technology Shanxi University

Local causal structure learning (LCS) efficiently identifies a set of direct neighbors of a specified variable from observational data. Additionally, it distinguishes direct causes and direct effects of this variable without learning the entire causal structure. While many LCS algorithms have been proposed, they do not consider the data privacy-preserving problem,which has attracted extensive attention from academia and industry. To address this issue, we propose a federated local causal structure learning (FedLCS) algorithm to learn local causal structures in privacy-preserving data in a federated ***, FedLCS introduces a layer-wise federated local skeleton learning algorithm to construct the local skeleton. Based on this skeleton, it introduces a federated local skeleton orientation algorithm and an extension-and-backtracking orientation algorithm to orient the edges. Finally, FedLCS uses a federated local extension-and-backtracking orientation algorithm to orient the remaining edges. Extensive experiments on benchmark, synthetic, and real datasets demonstrate that FedLCS can learn the local causal structure of a given variable in a federated setting.

关键词：