检索结果-内蒙古大学图书馆

$\mathrm{D}^{3}$T: Deep Denoising Dictionary Tensor for Hyperspectral Anomaly Detection

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2024年 18卷 3713-3727页

作者： Qiangqiang Shen Zonglin Liu Hanzhang Wang Yanhui Xu Yongyong Chen Yongsheng Liang School of Electronics and Information Engineering Harbin Institute of Technology Shenzhen China School of Science Harbin Institute of Technology Shenzhen China College of Electronic and Optical Engineering Nanjing University of Posts and Telecommunications Nanjing China School of Computer Science and Technology Harbin Institute of Technology Shenzhen China College of Big Data and Internet Shenzhen Technology University Shenzhen Guangdong China School of Electronics and Information Engineering Harbin Institute of Technology Shenzhen Guangdong China

Hyperspectral anomaly detection (HAD) approaches with tensor low-rank representation (TLRR) have shown engaging performance, which are capable of capturing abundant spectral and spatial information and, hence, extracting sparse anomalies from the low-rank background without ruining the intrinsic geometric structures of tensor data. However, existing HAD approaches construct the low-rank representation by a handcrafted dictionary that still contains the anomalies, resulting in an inferior representation. To this end, we propose a deep denoising dictionary tensor for hyperspectral anomaly detection ( $\mathrm{D}^{3}$ T), which can balance performance and interpretability by fusing two intuitive priors, i.e., low-rankness and deep denoising, into the dictionary tensor and coefficient tensor. In particular, the proposed $\mathrm{D}^{3}$ T first designs a tensor recovery module for dividing the low-rank background and sparse anomalies, where the background is optimized as a denoising dictionary tensor by the joint low-rankness and deep denoising. Following that, we build the TLRR model based on the denoising dictionary tensor to explore the latent representation of the input hyperspectral images, and hence, each background pixel is represented by the other clear background pixels in place of the anomaly pixels. Meanwhile, the coefficient tensor in TLRR is also optimized by the two priors to fully explore the spatial and spectral correlations in the background, and hence, the anomalies can be accurately detected. Numerous experimental results from various real-world datasets validate the effectiveness of our $\mathrm{D}^{3}$ T.

关键词： Tensors Dictionaries Noise reduction Anomaly detection Hyperspectral imaging Sparse matrices Noise Principal component analysis Feature extraction Electronic mail

来源：评论

学校读者我要写书评

暂无评论

Backdoor Attack Based on Privacy Inference against Federated Learning

Backdoor Attack Based on Privacy Inference against Federated...

引用

International Symposium on Autonomous Systems (ISAS)

作者： Duo Wu Lingguang Hao Bing Wei Kuangrong Hao Tao Han Langping He College of Information Science and Technology Donghua University Shanghai China College of Big Data and Artificial Intelligence China Telecom Research Institute Shanghai China

ISBN: (数字)9798350363173

ISBN: (纸本)9798350363180

As a pioneering distributed learning framework, federated learning (FL) has gained widespread adoption. It operates collaboratively among participants, with communication limited to sharing model parameters between the server and participants. However, FL is also more susceptible to active attacks from malicious insiders. Poisoned updates submitted by attackers can degrade the performance of the global model. Previous research only considered using the naive data of malicious clients for backdoor poisoning, and therefore achieved limited backdoor attack success rates. In this paper, we propose a novel Federated Backdoor Attack based on Privacy Inference (FBA-PI). Combining privacy inference based on generative adversarial networks (GAN), the attacker first infers sensitive information from victim participants and then injects backdoor triggers into the naive and generated data. Finally, malicious clients can embed backdoor semantics into the global model by participating in regular federated aggregation. Extensive experiments on the MNIST dataset verify the effectiveness of our proposed method.

关键词： Training Privacy data privacy Computer aided instruction Federated learning Distance learning Semantics

来源：评论

学校读者我要写书评

暂无评论

Fusing Pruned and Backdoored Models: Optimal Transport-based data-free Backdoor Mitigation

arXiv

引用

arXiv 2024年

作者： Lin, Weilin Liu, Li Li, Jianze Xiong, Hui The Hong Kong University of Science and Technology Guangzhou China Shenzhen Research Institute of Big Data China The Chinese University of Hong Kong Shenzhen China

Backdoor attacks present a serious security threat to deep neuron networks (DNNs). Although numerous effective defense techniques have been proposed in recent years, they inevitably rely on the availability of either clean or poisoned data. In contrast, data-free defense techniques have evolved slowly and still lag significantly in performance. To address this issue, different from the traditional approach of pruning followed by fine-tuning, we propose a novel data-free defense method named Optimal Transport-based Backdoor Repairing (OTBR) in this work. This method, based on our findings on neuron weight changes (NWCs) of random unlearning, uses optimal transport (OT)-based model fusion to combine the advantages of both pruned and backdoored models. Specifically, we first demonstrate our findings that the NWCs of random unlearning are positively correlated with those of poison unlearning. Based on this observation, we propose a random-unlearning NWC pruning technique to eliminate the backdoor effect and obtain a backdoor-free pruned model. Then, motivated by the OT-based model fusion, we propose the pruned-to-backdoored OT-based fusion technique, which fuses pruned and backdoored models to combine the advantages of both, resulting in a model that demonstrates high clean accuracy and a low attack success rate. To our knowledge, this is the first work to introduce OT and model fusion techniques to the backdoor defense. Extensive experiments show that our method successfully defends against all seven backdoor attacks across three benchmark datasets, outperforming both state-of-the-art (SOTA) data-free and data-dependent methods. © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

CLNX: Bridging Code and Natural Language for C/C++ Vulnerability-Contributing Commits Identification 39

CLNX: Bridging Code and Natural Language for C/C++ Vulnerabi...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Qin, Zeqing Wu, Yiwei Han, Lansheng School of Cyber Science and Engineering Huazhong University of Science and Technology Wuhan China Hubei Key Laboratory of Distributed System Security Hubei Engineering Research Center on Big Data Security China Wuhan JinYinHu Laboratory China

ISBN: (纸本)157735897X

Large Language Models (LLMs) have shown great promise in vulnerability identification. As C/C++ comprise half of the open-source Software (OSS) vulnerabilities over the past decade and updates in OSS mainly occur through commits, enhancing LLMs' ability to identify C/C++ Vulnerability-Contributing Commits (VCCs) is essential. However, current studies primarily focus on further pre-training LLMs on massive code datasets, which is resource-intensive and poses efficiency challenges. In this paper, we enhance the ability of BERT-based LLMs to identify C/C++ VCCs in a lightweight manner. We propose CodeLinguaNexus (CLNX) as a bridge facilitating communication between C/C++ programs and LLMs. Based on commits, CLNX efficiently converts the source code into a more natural representation while preserving key details. Specifically, CLNX first applies Structure-level Naturalization to decompose complex programs, followed by Token-level Naturalization to interpret complex symbols. We evaluate CLNX on public datasets of 25,872 C/C++ functions with their commits. The results demonstrate that CLNX substantially improves the ability of LLMs to detect C/C++ VCCs. Moreover, CLNX-equipped CodeBERT achieves new state-of-the-art performance and identifies 38 OSS vulnerabilities in the real world. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Open source software

来源：评论

学校读者我要写书评

暂无评论

PulseSync BP: A Neural Network Framework for Non-Contact Blood Pressure Estimation

PulseSync BP: A Neural Network Framework for Non-Contact Blo...

引用

Advances in data Engineering and Intelligent Computing Systems (ADICS), International Conference on

作者： Tanumay Ghosh Omisha Singal Namit Lodh S Suchitra Department of Data Science and Business Systems School of Computing SRM Institute of Science and Technology Kattankulathur Tamil Nadu India

ISBN: (数字)9798350364828

ISBN: (纸本)9798350364835

Using a cutting-edge neural network framework, “PulseSync BP” estimates blood pressure without contact. Using photoplethysmogram (PPG) signals and other physiological data, this novel model uses advanced signal processing. The suggested framework changes blood pressure monitoring by eliminating cuff-based measures. PulseSync BP's RNN architecture is carefully designed for optimal prediction accuracy. The algorithm effectively estimates blood pressure through extensive experimentation, demonstrating its potential for continuous healthcare applications. This framework's capacity to use various physiological cues shows its adaptability and versatility. PulseSync BP allows real-time, non-intrusive blood pressure monitoring, advancing individualized health tracking solutions. Cuffless blood pressure estimate is advanced here, demonstrating how artificial intelligence could improve healthcare accessibility and patient-centric monitoring. PulseSync BP and other revolutionary technologies can transform blood pressure monitoring in healthcare.

关键词： Neural networks Signal processing algorithms Estimation Signal processing Blood pressure Real-time systems Biomedical monitoring

来源：评论

学校读者我要写书评

暂无评论

Fine-Grained Visual Classification of Flue-Cured Tobacco Leaf Grades Based on Vein Information 10

Fine-Grained Visual Classification of Flue-Cured Tobacco Lea...

引用

10th International Conference on Systems and Informatics, ICSAI 2024

作者： Chen, Yuyang Lian, Changwei Chen, Xianyong Gong, Tao Peng, Xiaodong Chen, Fei College of Computer and Data Science Fuzhou University Fuzhou350108 China Fujian Provincial Company of China National Tobacco Corporation Fuzhou350001 China School of Mechanical and Automotive Engineering Fujian University of Technology Fuzhou350118 China

ISBN: (纸本)9798331510138

Fine-grained visual classification aims to categorize similar subcategories, with challenges lying in the high similarity among different subcategories and the high variability within the same category. Grading flue-cured tobacco leaves can be viewed as a fine-grained visual classification task. To improve the efficiency of tobacco leaf grading and minimize the subjectivity associated with manual grading, we propose a grading method leveraging vein information specific to flue-cured tobacco leaves. First, we designed a vein extraction module to extract the vein feature information of tobacco leaves. Then, we selected features beneficial for classification through a feature filtering module. In the main branch of the network, we designed a bidirectional fusion module to integrate multi-scale features with vein features. Additionally, by optimizing the loss function, we emphasize discriminative features while suppressing the influence of redundant information. Our proposed method achieved a classification accuracy of 77.9% on the 2023 Fujian Flue-cured Tobacco dataset, which contains six grades. On the 2024 Shaowu Flue-Cured Tobacco dataset, which includes nine grades of tobacco leaves, the classification accuracy reaches 96.7%. These results surpass all existing methods, demonstrating state-of-the-art results. © 2024 IEEE.

关键词： Cigarette filters

来源：评论

学校读者我要写书评

暂无评论

Swarm Robots Decentralized Control using Reinforcement Learning Stepwise Training Method in the Encircling Task

Swarm Robots Decentralized Control using Reinforcement Learn...

引用

2022 Global Conference on Robotics, Artificial Intelligence and Information technology, GCRAIT 2022

作者： Liu, Yang Wang, Xu Nie, Yiming Dai, Bin Zhu, Qi Shang, Erke College of Intelligence Science and Technology National University of Defence Technology Changsha China Xi'An Key Laboratory of Big Data and Intelligent Vision Xidian University Xi'an China National Innovation Institute of Defense Technology Beijing China

ISBN: (数字)9781665481922

ISBN: (纸本)9781665481922

We consider a special multiple-robot task, pursuit-evasion (PE) game, in which a swarm of hunting robots (Pursuers) attempts to encircle a target robot (Evader) in a two-dimensional plane with some fixed obstacles on it. Instead of the central control method, we suggest a fully decentralized pursuit strategy generated by the deep reinforcement learning (DRL) which has greatly advanced the research in decision-making field in recent years. While the DRL method in this encircling task faces a sparse reward and a nonconvex space challenge which makes convergence result unacceptable. In this article, we have clarified the definition of a successful encirclement by escaping angel. In order to overcome the inefficient exploration and premature convergence caused by the above challenges during the training process, we use a stepwise training method of an adaptive various reward function. All prosecutors use the same policy given by stepwise training method and the multi-agent deep deterministic policy gradient (MADDPG) algorithms. Simulations and comparisons are provided to validate the effectiveness of this work. © 2022 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

3-D Printed Multi-Layer Low-Profile Coupler based on Vertically Integrated CRLH Transmission Line

3-D Printed Multi-Layer Low-Profile Coupler based on Vertica...

引用

International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications (MAPE)

作者： Mengze Li Jing He Yang Yang Jiwei Lian MengMeng Li Dazhi Ding School of Microelectronics Nanjing University of Science and Technology Nanjing China School of Electrical and Data Engineering University of Technology Sydney NSW Australia

ISBN: (数字)9798331533922

ISBN: (纸本)9798331533939

This paper demonstrates several low-profile multilayer electronics fabricated using a novel 3D printing technology. This novel 3D printing technology can fabricate multilayer models with conductive and dielectric materials. In order to reduce the insertion loss and dimension of the electronics, a vertically integrated composite right/left handed (VI-CRLH) transmission line is proposed. And VI-CRLH couplers with good performance are design.

关键词： Microwave antennas Fabrication Wireless communication Solid modeling Power transmission lines Couplers Three-dimensional printing Nonhomogeneous media Propagation losses Scattering parameters

来源：评论

学校读者我要写书评

暂无评论

Flame Detection Algorithm Based on Improved YOLOv8

Flame Detection Algorithm Based on Improved YOLOv8

引用

Computer Vision, Image and Deep Learning (CVIDL), International Conference on

作者： Kun Huang Qian Hu Jianyong Pi Weichao Hu Wenhua Li Juanmin Wang State Key Laboratory of Public Big Data College of Computer Science and Technology Guizhou University Guiyang China Guizhou mobile information technology company Guiyang China

ISBN: (数字)9798350373820

ISBN: (纸本)9798350373837

Flame detection is a key technology in intelligent firefighting, using object detection methods to locate and count flames in videos, thereby enabling intelligent monitoring of fire incidents. This has significant research value in areas such as public safety and community intelligent governance. In complex scenes, due to the complicated backgrounds of flame targets, non-uniform sizes and shapes, and sensitivity to lighting, there’s a tendency for missed detections. This paper proposes an improved flame detection algorithm based on YOLOv8, named Fire-YOLOv8, addressing these issues. Firstly, a Tri-RepVGG module is introduced into the backbone network for better flame feature extraction without increasing computational costs, ensuring the timeliness of detection. Secondly, in the neck part of the YOLOv8 network, a Progressive Feature Pyramid Network (AFPN) structure is introduced to enhance the fusion of information between layers, solving the issues of information loss and degradation present in YOLOv8 feature fusion. Finally, based on the SE channel attention mechanism, an SE-Detect detection head is proposed to effectively mitigate the issues of flame detection sensitivity to lighting. Experimental results show that Fire-YOLOv8 achieves 74.5% mAP50 and 47.9% mAP50-95 on a self-made flame dataset Multi-Fire, representing improvements of 4.4% and 3.3% respectively over the original YOLOv8n, demonstrating strong competitive capabilities.

关键词： Degradation Sensitivity Shape Fires Lighting Feature extraction Real-time systems

来源：评论

学校读者我要写书评

暂无评论

Study on the Electromagnetic Properties and Microwave Absorption Properties of Co2si by Non-Magnetic Metal Cu and Mn Single Doping

SSRN

引用

SSRN 2024年

作者： Zou, Jiang Xie, Quan Chen, Jing Wang, Bo Zhang, Yu School of Physics and Electronic Science Zunyi Normal University Guizhou Zunyi563000 China Guizhou University School of Big Data and Information Engineering Institute of New Optoelectronic Materials and Technology Guiyang550025 China

In this study, Co2Si and Co2Si-doped powder materials with different Cu,Mn proportions were prepared by mechanical alloying and thermal sintering. The results showed that all the sample peaks obtained by XRD measurement were consistent with the standard card (ICDD: 98-005-2281) of *** Ms values corresponding to 2 wt%, 4 wt%, and 8 wt% Cu doping were 21.75 emu/g, 23.79 emu/g, and 25.79 emu/g, respectively. The saturation magnetization of Co2Si samples doped with 2 at%, 4 at% and 8 at% Mn was 8.26 emu/g, 20.33 emu/g and 14.65 emu/g, *** doping of Cu,Mn significantly improved the comprehensive absorption performance of Co2Si alloy, and the lowest reflection loss was obtained when Cu was doped with Co2Si alloy, with an RLmin value of -67.61 dB and an EAB of 2.40 GHz,and the RLmin value of Mn doped with 4 at% was -67.04 dB, and the EAB was 2.39 GHz. © 2024, The Authors. All rights reserved.

关键词： Silicon alloys

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：