检索结果-内蒙古大学图书馆

arXiv 2022年

作者： Yu, Yang Liu, Qi Wu, Likang Yu, Runlong Yu, Sanshi Lei Zhang, Zaixi Anhui Province Key Laboratory of Big Data Analysis and Application School of Computer Science and Technology University of Science and Technology of China China State Key Laboratory of Cognitive Intelligence China

Federated recommendation (FedRec) can train personalized recommenders without collecting user data, but the decentralized nature makes it susceptible to poisoning attacks. Most previous studies focus on the targeted attack to promote certain items, while the untargeted attack that aims to degrade the overall performance of the FedRec system remains less explored. In fact, untargeted attacks can disrupt the user experience and bring severe financial loss to the service provider. However, existing untargeted attack methods are either inapplicable or ineffective against FedRec systems. In this paper, we delve into the untargeted attack and its defense for FedRec systems. (i) We propose ClusterAttack, a novel untargeted attack method. It uploads poisonous gradients that converge the item embeddings into several dense clusters, which make the recommender generate similar scores for these items in the same cluster and perturb the ranking order. (ii) We propose a uniformity-based defense mechanism (UNION) to protect FedRec systems from such attacks. We design a contrastive learning task that regularizes the item embeddings toward a uniform distribution. Then the server filters out these malicious gradients by estimating the uniformity of updated item embeddings. Experiments on two public datasets show that ClusterAttack can effectively degrade the performance of FedRec systems while circumventing many defense methods, and UNION can improve the resistance of the system against various untargeted attacks, including our ClusterAttack. Copyright © 2022, The Authors. All rights reserved.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Backdoor Defense via Deconfounded Representation Learning

Backdoor Defense via Deconfounded Representation Learning

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Zaixi Zhang Qi Liu Zhicai Wang Zepu Lu Qingyong Hu Anhui Province Key Lab of Big Data Analysis and Application School of Computer Science and Technology University of Science and Technology of China State Key Laboratory of Cognitive Intelligence Hefei Anhui China University of Science and Technology of China Hong Kong University of Science and Technology

Deep neural networks (DNNs) are recently shown to be vulnerable to backdoor attacks, where attackers embed hidden backdoors in the DNN model by injecting a few poisoned examples into the training dataset. While extensive efforts have been made to detect and remove backdoors from backdoored DNNs, it is still not clear whether a backdoor-free clean model can be directly obtained from poisoned datasets. In this paper, we first construct a causal graph to model the generation process of poisoned data and find that the backdoor attack acts as the confounder, which brings spurious associations between the input images and target labels, making the model predictions less reliable. Inspired by the causal understanding, we propose the Causality-inspired Backdoor Defense (CBD), to learn deconfounded representations for reliable classification. Specifically, a backdoored model is intentionally trained to capture the confounding effects. The other clean model dedicates to capturing the desired causal effects by minimizing the mutual information with the confounding representations from the backdoored model and employing a sample-wise re-weighting scheme. Extensive experiments on multiple benchmark datasets against 6 state-of-the-art attacks verify that our proposed defense method is effective in reducing backdoor threats while maintaining high accuracy in predicting benign samples. Further analysis shows that CBD can also resist potential adaptive attacks. The code is available at https://***/zaixizhang/CBD.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Tibetan Ancient Uchen Text Line Dataset for OCR

A Tibetan Ancient Uchen Text Line Dataset for OCR

引用

Image Processing, computer Vision and Machine Learning (ICICML), International Conference on

作者： Zhuome Gongqu Peng Luo Dongzhou Jiayang Jia Cairang Jiacuo Cizhen Dongzhu Renqing The College of Computer Qinghai Normal University Xining China The State Key Laboratory of Tibetan Intelligent Information Processing and Application Xining China Tibetan Information Processing and Machine Translation Key Laboratory of Qinghai Province Xining China The College of Information Science and Technology Tibet University Lhasa China

ISBN: (数字)9798350355413

ISBN: (纸本)9798350355420

Characters written by hand that represent distinct writing styles are referred to as handwritten text; these are especially common in ancient texts. In the field of digitizing ancient literature, Tibetan ancient document image recognition is a focus of both domestic and international research. However, technological advancement in this field has been hampered by the absence of a specific recognition dataset for image recognition systems of Tibetan ancient documents. This paper addresses this by proposing a dataset for Uchen style text line recognition in Tibetan ancient documents. This dataset includes basic synthetic images of multi-style Uchen text lines as well as real-world images of multi-format Tibetan Uchen text lines that have been manually annotated. Cross-Validation and several rounds of annotation by various people guarantee the quality of the data. Additionally, three popular text line recognition models are used to assess the dataset and validate its usability.

关键词： Hands computer vision Image recognition Text recognition Annotations Computational modeling Optical character recognition Machine learning Writing Usability

来源：评论

学校读者我要写书评

暂无评论

Reliability of gold wire leads for MEMS gyroscopes under a thermo-mechanical coupling field

Reliability of gold wire leads for MEMS gyroscopes under a t...

引用

IEEE International Conference on Nano/Micro Engineered and Molecular Systems (NEMS)

作者： Yingyu Xu Chunhua He Qinwen Huang Guizhen Yan School of Computer Guangdong University of Technology Guangzhou China Science and Technology on Reliability Physics and Application Technology of Electronic Component Laboratory China Electronic Product Reliability and Environmental Testing Research Institute Guangzhou China National Key Laboratory of Science and Technology on Micro/Nano Fabrication Institute of Microelectronics Peking University Beijing China

ISBN: (数字)9798350359831

ISBN: (纸本)9798350359848

In this study, we investigate the vibration characteristics and failure modes of gold wire leads in MEMS gyroscopes under high-temperature and thermal-shock conditions through simulation and experimental analysis. The simulation results indicate a gradual degradation in the intrinsic frequencies of the leads as the temperature rises from -40 °C to 120 °C. The simulation results are verified by testing the intrinsic frequency with a percentage error of less than 2%. The average temperature drift of the intrinsic frequency is approximately - 8.82 Hz/°C. When comparing the intrinsic frequency changes before and after burn-in and thermal-shock tests, it is evident that the mechanical degradation of the gold wire leads is limited under the severe thermo-mechanical coupling field, with frequency fluctuations ranging from 0.01% to 17%. There is no occurrence of a lead fracture or solder joint failure. The main failure modes are lead collapse and short-circuit failure due to the fatigue of long leads. Minimizing the span and height of the leads can reduce the risk of fatigue and short-circuit failures.

关键词： Micromechanical devices Couplings Gold Simulation Thermomechanical processes Fatigue Gyroscopes

来源：评论

学校读者我要写书评

暂无评论

MambaRF: A Bi-directional Mamba Structure for Radio Frequency Signal Classification of Unmanned Aerial Vehicle

MambaRF: A Bi-directional Mamba Structure for Radio Frequenc...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Mufeng Yao Chao Liu Lexu Xie Mingmin Chi School of Computer Science Shanghai Key Laboratory of Data Science Fudan University Shanghai China School of Computer Science Fudan University Zhengzhou Zhongke Institute of Integrated Circuit and System Application Shanghai China Henan Institute of Advanced Technology Zhengzhou University Zhengzhou China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

The rapid development of Unmanned Aerial Vehicle (UAV) technology has facilitated the widespread use of UAVs in daily life, this advancement brings huge regulatory demand for UAVs. Automatic identification of UAVs using radio frequency (RF) signals can effectively reduce regulatory costs. Previous studies first extracts spectrogram features from raw RF signals using short-time Fourier transform, and then use neural networks to classify the spectrograms. These studies are mainly based on local convolution, ignoring the long-range dependence of the spectrograms, and thus are limited in terms of UAV classification accuracy. In addition, these works also focus only on classifying UAV types without identifying flight states (e.g., hovering, flying, and switch on). To this end, we propose MambaRF, which employs a bi-directional selective scanning mechanism to capture long-range information of spectrograms from two different directions. MambaRF also jointly classifies UAV types and flight states by using two decoupled classification heads, which has not been addressed in previous studies. Experiments on two publicly available datasets show that our proposed MambaRF outperforms VMamba and other previous studies in both type classification and flight state classification.

关键词： Accuracy Convolution RF signals Bidirectional control Switches Autonomous aerial vehicles Feature extraction Speech processing Spectrogram Radiofrequency identification

来源：评论

学校读者我要写书评

暂无评论

Research on Waxiang Dialect Recognition Based on Wav2vec2.0 Pre-Training and Transfer Learning

Research on Waxiang Dialect Recognition Based on Wav2vec2.0 ...

引用

Computing, Robotics and System Sciences (ICRSS), International Conference on

作者： Mixuan Deng Pengju Tang Yanxia Gao Huixuan Li Guodong Zhen Kaifeng Fu Xiaotian Wu College of Computer and Artificial Intelligence(Software College) Huaihua University Huaihua China Key Laboratory of Wuling-Mountain Health Big Data Intelligent Processing and Application in Hunan Province Universities Key Laboratory of Intelligent Control Technology for Wuling-Mountain Ecological Agriculture in Hunan Province Huaihua China

ISBN: (数字)9798331511296

ISBN: (纸本)9798331511302

As a unique local dialect, Waxiang dialect carries rich regional culture and historical information. However, with the acceleration of modernization and the increase of population mobility, the inheritance and protection of Waxiang dialect are facing severe challenges. By constructing the ‘Xuewatong’ APP- Waxiang dialect digital protection platform, based on cloud storage technology, digital resource display, multimedia integration and other technologies, and providing learning resources through big data analysis of user needs, it can effectively promote the digital recording and preservation of Waxiang dialect language resources, and significantly improve the dissemination and educational effect of Waxiang dialect culture, showing the great potential of digital technology in dialect protection.

关键词： Technological innovation Cloud computing Transfer learning Media Information age Digital recording Global communication Cultural differences Protection Robots

来源：评论

学校读者我要写书评

暂无评论

A Single-Loop Narrow-Band Force Rebalance Control Method with Temperature Self-Compensation for a MEMS Gyroscope

A Single-Loop Narrow-Band Force Rebalance Control Method wit...

引用

IEEE International Conference on Micro Electro Mechanical Systems

作者： Chunhua He Yingyu Xu Heng Wu Qinwen Huang Qiancheng Zhao Guizhen Yan School of Computer Guangdong University of Technology China Science and Technology on Reliability Physics and Application Technology of Electronic Component Laboratory CEPREI China School of Automation Guangdong University of Technology China School of Integrated Circuits Peking University Beijing China National Key Lab of Micro/Nano Fabrication Technology Beijing China

ISBN: (数字)9798331508890

ISBN: (纸本)9798331508906

This paper presents a novel single-loop narrow-band force rebalance control method with temperature self-compensation for a MEMS gyroscope. The narrow-bandwidth force rebalance control for the sense mode is achieved using a simple bandpass filter (BPF) and without modulation or demodulation. The center frequency of the BPF is adjusted automatically to match the resonant frequency of the sense mode. Meanwhile, its gain is tuned based on the real-time quality factor of the sense mode. Experimental results demonstrate that the bias drift of the gyroscope with self-compensation is reduced from 1320 deg/h to 16 deg/h, while the scale factor drift is reduced from 0.616 mV/deg/s to 0.075 mV/deg/s.

关键词： Micromechanical devices Band-pass filters Temperature sensors Q-factor Force Resonant frequency Real-time systems Gyroscopes Temperature control Phase locked loops

来源：评论

学校读者我要写书评

暂无评论

Intelligent Food Identification System Using Improved YOLOv8

Intelligent Food Identification System Using Improved YOLOv8

引用

2024 International Symposium on computer Science and Educational technology (ISCSET)

作者： Francia Nitiavintokana Md Abdur Rouf Xiaoyu Yu Haibin Wu Aili Wang Yuji Iwahori Heilongjiang Province Key Laboratory of Laser Spectroscopy Technology and Application Harbin University of Science and Technology Harbin China College of Electron and Information University of Electronic Science and Technology of China Zhongshan Institute Zhongshan China Department of Computer Science Chubu University Aichi Japan

ISBN: (数字)9798350390803

ISBN: (纸本)9798350390810

The field of computer vision has witnessed significant advances in object detection with real-world applications ranging from autonomous driving to surveillance systems. Due to its importance in a number of areas, such as dietary evaluation, food industry automation, and health monitoring, food identification has attracted a lot of interest in this context. The suggested approach makes use of the capabilities of two cutting-edge object recognition models, YOLOv8. Because of its speed and accuracy, YOLOv8 is a great option for real-time food detection in a variety of settings. Our most recent YOLOv8 model is called BiTrains-YOLOv8. Supervision techniques are applied to improve the combined model's performance even more. This increases overall detection accuracy and helps the algorithm to more effectively generalize to novel food items. Based on the detection results, 96.7% of an average of detection accuracy, 96.4% for precision, 95.7 % for recall, and 96.04 % for F1-Score were obtained. The results show that the combination of YOLOv8 and BiTrains-YOLOv8 improved model techniques, significantly improved the performance of food detection, outperforming the previous methods in terms of accuracy, speed and robustness. Furthermore, the paper analyzes potential real-world applications of the system, such as calorie counting, dietary assessment and smart kitchen automation. Implications of advanced intelligent food detection in healthcare, the culinary industry and personalized nutrition are also discussed. In conclusion, the results provide valuable insights for pushing the boundaries of food recognition technology, with promising applications in various fields that promote healthy lifestyles and improve efficiency in food-related processes.

关键词： Training Industries Accuracy Automation Surveillance Object detection Medical services Robustness Real-time systems Object recognition

来源：评论

学校读者我要写书评

暂无评论

A New Fourth-Order Sparse Array Generator Based on Sum-Difference Co-Array Analysis

A New Fourth-Order Sparse Array Generator Based on Sum-Diffe...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Haodong Guo Hua Chen Hongguang Lin Wei Liu Qing Shen Gang Wang Faculty of Electrical Engineering and Computer Science Ningbo University Ningbo China Zhejiang Key Laboratory of Mobile Network Application Technology Ningbo China School of Electronic Engineering and Computer Science Queen Mary University of London London UK School of Information and Electronics Beijing Institute of Technology China

In this paper, based on sum-difference co-array analysis, a new fourth-order sparse array called sum-difference-FODC (SD-FODC) is proposed, allowing the construction of a fourth-order DCA with long consecutive lags using the continuous segments in the second-order DCA and SCA of the original array. It has a closed-form expression for sensor positions that can be generated by two arbitrary nonuniform linear arrays (NLAs) called generators. If the second-order SCA and DCA of the generators have long consecutive segments, the designed fourth-order sparse array can achieve a large number of uDOFs. Numerical results are provided to demonstrate the superior performance of the proposed design.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hierarchical multi-granularity classification based on bidirectional knowledge transfer

引用

Multimedia Systems 2024年第4期30卷 207-207页

作者： Jiang, Juan Yang, Jingmin Zhang, Wenjie Zhang, Hongbin School of Computer Science Minnan Normal University Fujian Zhangzhou363000 China Key Laboratory of Data Science and Intelligence Application Fujian Province University Fujian Zhangzhou363000 China Department of Electronic Engineering National Taipei University of Technology Taipei China

Hierarchical multi-granularity classification is the task of classifying objects according to multiple levels or granularities. The class hierarchy is vital side information for hierarchical multi-granularity classification. The existing hierarchical multi-granularity classification research utilizes class hierarchy to classify from coarse to fine or fine to coarse. Although these methods are effective in many cases, there are still two issues: (1) multi-task learning for hierarchical multi-granularity classification leads to decreased classification performance;(2) class hierarchy transfer learning is prone to error propagation. In this paper, we propose a bidirectional knowledge transfer model framework to address these issues. Firstly, we improve classification performance through data augmentation. Specifically, by learning the similarity between the original image and the enhanced image, better learn discriminative features, which is beneficial for subsequent classification. Secondly, using class hierarchy trees, we propose reverse hierarchical knowledge transfer to correct some errors in forward hierarchical propagation and improve hierarchical consistency. In addition, we also construct a hierarchical network that adds features from coarse-grained levels to fine-grained levels. The experimental results on six datasets with different class hierarchies demonstrate the effectiveness and superiority of the proposed model. Especially on the CUB-200-2011 and Cifar-100 datasets, our model improved classification accuracy by 3.61% and 4.17% compared to the suboptimal model. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Classification (of 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：