检索结果-内蒙古大学图书馆

IEEE International Conference on Signal and Image Processing (ICSIP)

作者： Yacong Yu Lei Zhang Liquan Chen Zhongyuan Qin Key Laboratory of Computer Networking Technology of Jiangsu Province School of Cyber Science and Engineering Southeast University Nanjing China

ISBN: (纸本)9781665446006

Adversarial sample attacks seriously threaten the security and robustness of deep learning models. There are three problems in state-of-the-art adversarial sample generation schemes: the gradient update step size needs to be manually selected, inaccurate gradient update direction and uncontrollable times of iterations. In order to solve these problems, Root Mean Square Prop optimization algorithm (RMSProp) is proposed, which is integrated with IFGSM and IFGM. This algorithm can be easily extended to other attacks, and to a certain extent it can alleviate the trade-off between white box attacks and deliverability. The algorithm proposed in this paper can generate non-targeted adversarial samples more efficiently and quickly. Experiments show that it can generate effective and robust adversarial samples against current mainstream convolutional neural network (CNN).

关键词： Deep learning Image processing Conferences Robustness Convolutional neural networks Security Iterative methods

来源：评论

学校读者我要写书评

暂无评论

COUNTING RANDOM k-SAT NEAR THE SATISFIABILITY THRESHOLD

arXiv

引用

arXiv 2024年

作者： Chen, Zongchen Lonkar, Aditya Wang, Chunyang Yang, Kuan Yin, Yitong School of Computer Science Georgia Institute of Technology AtlantaGA United States State Key Laboratory for Novel Software Technology New Cornerstone Science Laboratory Nanjing University 163 Xianlin Avenue Jiangsu Province Nanjing210023 China John Hopcroft Center for Computer Science Shanghai Jiao Tong University Shanghai China

We present efficient counting and sampling algorithms for random k-SAT when the clause density satisfies α ≤ 2k poly(k) . In particular, the exponential term 2k matches the satisfiability threshold Θ(2k) for the existence of a solution and the (conjectured) algorithmic threshold 2k (ln k)/k for efficiently finding a solution. Previously, the best-known counting and sampling algorithms required far more restricted densities α 2k/3 [HWY23]. Notably, our result goes beyond the lower bound d 2k/2 for worst-case k-SAT with bounded-degree d [BGG+19], showing that for counting and sampling, the average-case random k-SAT model is computationally much easier than the worst-case model. At the heart of our approach is a new refined analysis of the recent novel coupling procedure by [WY24], utilizing the structural properties of random constraint satisfaction problems (CSPs). Crucially, our analysis avoids reliance on the 2-tree structure used in prior works, which cannot extend beyond the worst-case threshold 2k/2. Instead, we employ a witness tree similar to that used in the analysis of the Moser-Tardos algorithm [MT10] for the Lovász Local lemma, which may be of independent interest. Our new analysis provides a universal framework for efficient counting and sampling for random atomic CSPs, including, for example, random hypergraph colorings. At the same time, it immediately implies as corollaries several structural and probabilistic properties of random CSPs that have been widely studied but rarely justified, including replica symmetry and non-reconstruction. Copyright © 2024, The Authors. All rights reserved.

关键词： Trees (mathematics)

来源：评论

学校读者我要写书评

暂无评论

Towards Optimal Customized Architecture for Heterogeneous Federated Learning with Contrastive Cloud-Edge Model Decoupling

arXiv

引用

arXiv 2024年

作者： Chen, Xingyan Du, Tian Wang, Mu Gu, Tiancheng Zhao, Yu Kou, Gang Xu, Changqiao Wu, Dapeng Oliver Financial Intelligence and Financial Engineering Key Laboratory of Sichuan Province Institute of Digital Economy and Interdisciplinary Science Innovation School of Computer and Artificial Intelligence Southwestern University of Finance and Economics Chengdu611130 China The State Key Laboratory of Networking and Switching Technology Beijing University of Posts and Telecommunications Beijing100876 China The Data Engineering at the Department of Computer Science City University of Hong Kong Hong Kong

Federated learning, as a promising distributed learning paradigm, enables collaborative training of a global model across multiple network edge clients without the need for central data collecting. However, the heterogeneity of edge data distribution drags the model towards the local minima, which can be distant from the global optimum. Such heterogeneity often leads to slow convergence and substantial communication overhead. To address these issues, we propose a novel federated learning framework called FedCMD, a model decoupling tailored to the Cloud-edge supported federated learning that separates deep neural networks into a body for capturing shared representations in Cloud and a personalized head for migrating data heterogeneity. Our motivation is that, by the deep investigation of the performance of selecting different neural network layers as the personalized head, we found rigidly assigning the last layer as the personalized head in current studies is not always optimal. Instead, it is necessary to dynamically select the personalized layer that maximizes the training performance by taking the representation difference between neighbor layers into account. To find the optimal personalized layer, we utilize the low-dimensional representation of each layer to contrast feature distribution transfer and introduce a Wasserstein-based layer selection method, aimed at identifying the best-match layer for personalization. Additionally, a weighted global aggregation algorithm is proposed based on the selected personalized layer for the practical application of FedCMD. Extensive experiments on ten benchmarks demonstrate the efficiency and superior performance of our solution compared with nine state-of-the-art solutions. All code and results are available at https://***/elegy112138/FedCMD. Copyright © 2024, The Authors. All rights reserved.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

Adaptive optical quantitative phase imaging based on annular illumination Fourier ptychographic microscopy

引用

PhotoniX 2022年第1期3卷 101-115页

作者： Yefeng Shu Jiasong Sun Jiaming Lyu Yao Fan Ning Zhou Ran Ye Guoan Zheng Qian Chen Chao Zuo Smart Computational Imaging Laboratory(SCILab) School of Electronic and Optical EngineeringNanjing University of Science and Technolog210094 NanjingJiangsu ProvincePeople’s Republic of China Smart Computational Imaging Research Institute(SCIRI)of Nanjing University of Science and Technology 210019 NanjingJiangsu ProvincePeople’s Republic of China Jiangsu Key Laboratory of Spectral Imaging Intelligent Sense 210094 NanjingJiangsu ProvincePeople’s Republic of China Terahertz Technology Innovation Research Institute University of Shanghai for Science and Technology200093 ShanghaiPeople’s Republic of China School of Computer and Electronic Information Nanjing Normal University210023 NanjingJiangsu ProvincePeople’s Republic of China Department of Biomedical Engineering University of Connecticut06269 StorrsConnecticutUSA

Quantitative phase imaging(QPI)has emerged as a valuable tool for biomedical research thanks to its unique capabilities for quantifying optical thickness variation of living cells and *** many QPI methods,Fourier ptychographic microscopy(FPM)allows long-term label-free observation and quantitative analysis of large cell populations without compromising spatial and temporal ***,high spatio-temporal resolution imaging over a long-time scale(from hours to days)remains a critical challenge:optically inhomogeneous structure of biological specimens as well as mechanical perturbations and thermal fluctuations of the microscope body all result in time-varying aberration and focus drifts,significantly degrading the imaging performance for long-term ***,the aberrations are sample-and environmentdependent,and cannot be compensated by a fixed optical design,thus necessitating rapid dynamic correction in the imaging ***,we report an adaptive optical QPI method based on annular illumination *** this method,the annular matched illumination configuration(i.e.,the illumination numerical aperture(NA)strictly equals to the objective NA),which is the key for recovering low-frequency phase information,is further utilized for the accurate imaging aberration *** using only 6 low-resolution images captured with 6 different illumination angles matching the NA of a 10x,0.4 NA objective,we recover high-resolution quantitative phase images(synthetic NA of 0.8)and characterize the aberrations in real time,restoring the optimum resolution of the system *** our method to live-cell imaging,we achieve diffraction-limited performance(full-pitch resolution of 655 nm at a wavelength of 525 nm)across a wide field of view(1.77mm2)over an extended period of time.

关键词： Quantitative phase imaging Fourier ptychographic microscopy Adaptive optics Long-term imaging

来源：评论

学校读者我要写书评

暂无评论

A CPK-Based Identity Authentication Scheme for IoT

引用

computer Systems Science & Engineering 2022年第3期40卷 1217-1231页

作者： Mingming Zhang Jiaming Mao Yuanyuan Ma Liangjie Xu Chuanjun Wang Ran Zhao Zhihao Li Lu Chen Wenbing Zhao State Grid Jiangsu Electric Power Co. Ltd.Information&Telecommunication BranchNanjing210008China State Grid Key Laboratory of Information&Network Security Institute of Information and CommunicationGlobal Energy Interconnection Research InstituteNanjing210003China Anhui Jiyuan Software Co.Ltd SGITGHefei230088China Engineering Research Center of Post Big Data Technology and Application of Jiangsu Province Research and Development Center of Post Industry Technology of the State Posts Bureau(Internet of Things Technology)Engineering Research Center of Broadband Wireless Communication Technology of the Ministry of EducationNanjing University of Posts and TelecommunicationsNanjing210003China Department of Electrical Engineering and Computer Science Cleveland State UniversityCleveland44115USA

As the power Internet of Things(IoT)enters the security construction stage,the massive use of perception layer devices urgently requires an identity authentication scheme that considers both security and *** existing public key infrastructure(PKI)-based security authentication scheme is currently difficult to apply in many terminals in *** key distribution and management costs are high,which hinders the development of power IoT security *** Public key(CPK)technology uses a small number of seeds to generate unlimited public *** is very suitable for identity authentication in the power Internet of *** this paper,we propose a novel identity authentication scheme for power *** scheme combines the physical unclonable function(PUF)with improved CPK technology to achieve mutual identity authentication between power IoT terminals and *** proposed scheme does not require third-party authentication and improves the security of identity authentication for power ***,the scheme reduces the resource consumption of power IoT *** improved CPK algorithm solves the key collision problem,and the third party only needs to save the private key and the public key *** results show that the amount of storage resources occupied in our scheme is *** proposed scheme is more suitable for the power IoT.

关键词： Power Internet of Things terminal equipment identity authentication CPK PUF

来源：评论

学校读者我要写书评

暂无评论

CADEC: A Combinatorial Auction for Dynamic Distributed DNN Inference Scheduling in Edge-Cloud Networks

引用

IEEE Transactions on Mobile Computing 2025年

作者： Xu, Xiaolong Hu, Yuhao Cui, Guangming Qi, Lianyong Dou, Wanchun Nanjing University of Information Science and Technology School of Software Nanjing210044 China Nanjing University of Information Science and Technology Jiangsu Province Engineering Research Center of Advanced Computing and Intelligent Services School of Software China Nanjing University State Key Laboratory for Novel Software Technology China College of Computer science and technology Qingdao China Georgia State University Department of Computer Science United States

Deep Neural Network (DNN) Inference, as a key enabler of intelligent applications, is often computation-intensive and latency-sensitive. Combining the advantages of cloud computing (abundant computing resources) and edge computing (fast transmission), edge-cloud collaborative DNN inference is a powerful solution to these problems. However, in edge-cloud networks with heterogeneous resources, how to obtain reasonable decisions on server selection, model partition and resource allocation for efficient distributed DNN inference is a hard challenge. Furthermore, it is non-trivial to design suitable resource prices to maximize the social welfare. These challenges even escalate in dynamic edge-cloud networks where decisions should be generated as soon as each user arrives without future information. Therefore, we design a combinatorial auction for dynamic distributed DNN inference scheduling, named CADEC. CADEC first constructs a bid set for each user based on convex optimization theory for optimal solution searching. Next, prices of resources in the edge-cloud network are adjusted according to changes in supply-demand relationship, and whether to admit the request of each user is decided. Finally, the dynamic distributed inference scheduling decisions are generated through the primal-dual algorithm to maximize the social welfare. Theoretical analysis shows the good competitive ratio and polynomial time complexity of CADEC. Results of simulation experiments present that CADEC improves social welfare by up to 46% compared with state-of-the-art distributed DNN inference schemes. © 2002-2012 IEEE.

关键词： Convex optimization

来源：评论

学校读者我要写书评

暂无评论

NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification 39

NLGT: Neighborhood-based and Label-enhanced Graph Transforme...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Xu, Xiaolong Zhou, Yibo Xiang, Haolong Li, Xiaoyong Zhang, Xuyun Qi, Lianyong Dou, Wanchun School of Software Nanjing University of Information Science and Technology China Yunnan Key Laboratory of Service Computing Yunan University of Finance and Economics China Jiangsu Province Engineering Research Center of Advanced Computing and Intelligent Services China Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology China College of Meteorology and Oceanography National University of Defense Technology China School of Computing Macquarie University Australia College of Computer Science and Technology China University of Petroleum (East China) China State Key Laboratory for Novel Software Technology Nanjing University China

ISBN: (纸本)157735897X

Graph Neural Networks (GNNs) are widely applied on graph-level tasks, such as node classification, link prediction and graph generation. Existing GNNs mostly adopt a message-passing mechanism to aggregate node information with their neighbors, which often makes node information similar after rounds of aggregations and leads to oversmoothing. Although recent works have made improvements by combining different message aggregation methods or introducing semantic encodings as priors, these message-passing based GNNs still fail to combat oversmoothing after multiple iterations of node aggregation. Besides, the feature extraction ability of these methods is restricted because of the graph sparsity that hinders the aggregation of node information. To deal with the above two issues, we propose Neighborhood-based and Label-enhanced Graph Transformer (NLGT), a novel and effective framework for graph learning. Specifically, we present a label-enhanced feature fusion mechanism that integrate the shallow node features and label embeddings as enhanced features. Moreover, we design a neighborhood-based mask attention mechanism to alleviate the negative effects caused by the sparsity of the graph. In the predicting stage, we aggregate the prediction results from multiple sampled sub-graphs and apply voting mechanisms to enhance the accuracy and robustness of our framework. Finally, extensive experiments are conducted on four open benchmark datasets, which demonstrate the effectiveness and robustness of our proposed framework compared with existing state-of-the-art methods. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Clustering Interval and Triangular Granular Data: Modeling, Execution, and Assessment

引用

IEEE Transactions on Neural Networks and Learning Systems 2024年第6期36卷 10000-10014页

作者： Yiming Tang Wenbin Wu Witold Pedrycz Jianwei Gao Xianghui Hu Zhaohong Deng Rui Chen Anhui Province Key Laboratory of Affective Computing and Advanced Intelligent Machine School of Computer and Information Hefei University of Technology Hefei China Department of Electrical and Computer Engineering University of Alberta Edmonton AB Canada School of Computer and Information Hefei University of Technology Hefei China Faculty of Automatic Control Electronics and Computer Science Silesian University of Tecghnology Gliwice Poland Research Center of Performance and Productivity Analysis Istinye University Istanbul TÃrkiye School of Computer Science and Engineering Southeast University Nanjing China School of Artificial Intelligence and Computer Jiangnan University Wuxi Jiangsu China

In current granular clustering algorithms, numeric representatives were selected by users or an ordinary strategy, which seemed simple; meanwhile, weight settings for granular data could not adequately express their structural characteristics. Aiming at these problems, in this study, a new scheme called a granular weighted kernel fuzzy clustering (GWKFC) algorithm is put forward. We propose the representative selection and granularity generation (RSGG) algorithm enlightened by the density peak clustering (DPC) algorithm. We build interval and triangular granular data on the strength of numeric representatives obtained by RSGG under the principle of justifiable granularity (PJG), in which we establish some combinations of functions and boundary constraints and prove their properties. Furthermore, we present a novel distance formula via the kernel function for granular data and design new weights to affect the coverage and specificity of granular data. In addition, based upon these factors, we come up with the GWKFC algorithm of granular clustering, and its performance with different granularity is assessed. To sum up, a macro framework involving granular modeling, granular clustering, and assessment has been set up. Lastly, the GWKFC algorithm and ten other granular clustering algorithms are compared by experiments on some artificial and UCI datasets together with datasets with large data or those of high dimensionality. It is found that the GWKFC algorithm can provide better granular clustering results by contrast with other algorithms. The originality is embodied as follows. First, we improve the previous density radius and present the RSGG algorithm to acquire numeric representatives. Second, we propose a new strategy to determine granular data boundaries and further obtain novel weights enlightened by the idea of volume. Lastly, we employ the kernel function to calculate the distance between granular data, which has a stronger spatial division ability than the pr

关键词： Clustering algorithms Data models Computational modeling Numerical models Kernel Phase change materials Granular computing Uncertainty Fuzzy sets Noise

来源：评论

学校读者我要写书评

暂无评论

Background-Driven and Foreground-Refined Network for Weakly Supervised Change Detection

引用

IEEE Transactions on Geoscience and Remote Sensing 2025年 63卷

作者： Liu, Fang Zhang, Peng Liu, Jia Yang, Jingxiang Tang, Xu Xiao, Liang Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing210094 China Jiangsu Province Engineering Research Center of Airborne Detecting and Intelligent Perceptive Technology Nanjing210049 China Nanjing University of Science and Technology School of Computer Science and Engineering Key Laboratory of Analysis of Mathematical Theory and Modeling of Complex Systems Ministry of Industry and Information Technology Nanjing210094 China Xidian University School of Artificial Intelligence Xi'an710071 China

Change detection (CD) in remote sensing aims to reveal meaningful surface changes and has been flourishing in recent years. Compared with fully supervised methods based on pixel-level labels, image-level labels are easy to acquire, which reduces manual labor to a large extent. However, image-level labels lack spatial-and-shape information while containing the least semantic information, which poses a great challenge to the weakly supervised CD task. Motivated by the prior that bi-temporal images have background semantic consistency, we propose a background-driven and foreground-refined network (BDFR-Net) to ameliorate the above problem. Specifically, there are two key components in the proposed method: background-driven reconstruction (BDR) with image-level supervision and foreground-refined learning (FRL) with affinity learning. The former generates changed regions of foreground and background separation, which activates the foreground from image-level supervision and constrains the foreground by maintaining spatial and semantic consistency in background regions. The latter introduces complementary fusion and label adaption (CFLA) strategies to further refine the foreground, which can mine complementary information from foreground sequences and suppress false activations. In addition, affinity learning is proposed to stabilize and supervise the above process. Complementary relationships between the foreground and the background are fully utilized. Tested on two popular CD datasets, the results demonstrate that our proposed BDFR-Net produces completely changed regions with clear boundaries and outperforms state-of-the-art weakly supervised methods. © 1980-2012 IEEE.

关键词： Change detection

来源：评论

学校读者我要写书评

暂无评论

Enhancing Wi-Fi Device Authentication Protocol Leveraging Channel State Information 14th

Enhancing Wi-Fi Device Authentication Protocol Leveraging Ch...

引用

14th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2021

作者： Chen, Bing Song, Yubo Wu, Tianqi Zheng, Tianyu Chen, Hongyuan Wang, Junbo Li, Tao School of Cyber Science and Engineering Key Laboratory of Computer Network Technology of Jiangsu Province Southeast University Nanjing China School of Information Science and Engineering Southeast University Nanjing China Purple Mountain Laboratories Nanjing China

ISBN: (纸本)9783030898137

Wi-Fi device authentication is crucial for defending against impersonation attacks and information forgery attacks. Most of the existing authentication technologies rely on complex cryptographic algorithms. However, they cannot be supported well on the devices with limited hardware resources. A fine-grained device authentication technology based on channel state information (CSI) provides a non-cryptographic method, which uses the fingerprint extracted from CSI for authentication since CSI can uniquely identify the device in a limited time. But maintaining a fingerprint database for fingerprint matching is a challenging work. Firstly, the fingerprints extracted from the CSI are time-sensitive, which means that the fingerprint database must be updated in real time;Secondly, the authentication device may collect false fingerprints under the attack of identity-based attackers, which means that the authenticity of the fingerprint used to update the database must be checked. In this paper, we propose an enhancing Wi-Fi device authentication protocol based on CSI to implement the fingerprint database update and the device authentication. We provide a viable method of database update and an authentication algorithm based on Local Outlier Factor (LOF). We also present a complete authentication process. In addition, we evaluate the performance of our CSI-based authentication algorithm and database updating method. The experiments showed that the accuracy of the authentication algorithm is up to 97.1% and our database updating method can help the system maintain high accuracy. © 2021, ICST Institute for computer Sciences, Social Informatics and Telecommunications Engineering.

关键词： Channel state information

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：