检索结果-内蒙古大学图书馆

Semantic-Aware Pseudo-Labeling for Unsupervised Meta-Learning

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年第7期47卷 5475-5488页

作者： Ouyang, Tianran Dong, Xingping Ye, Mang Du, Bo Shao, Ling Shen, Jianbing Wuhan University School of Computer Science National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan430072 China University of Chinese Academy of Sciences UCAS-Terminus AI Lab Beijing101408 China University of Macau State Key Laboratory of Internet of Things for Smart City Department of Computer and Information Science 314100 China

In unsupervised meta-learning, the clustering-based pseudo-labeling approach is an attractive framework, since it is model-agnostic, allowing it to synergize with supervised algorithms to learn from unlabeled data. However, the pseudo-labels suffer from clustering noise and semantic chaos problems, further impacting the effectiveness of meta-learning. In this paper, we analyze and optimize the pseudo-labeling process, including encoding and clustering, aiming to generate semantic-like pseudo-labels to narrow the gap between unsupervised and supervised meta-learning. First, during the encoding, we observe that the embedding space of existing methods lacks clustering-friendly properties, which is the primary reason for clustering noise. To address this issue, we minimize the inter-to-intra-class similarity ratio to generate clustering-friendly embedding features and validate our approach through comprehensive experiments. Then, during the clustering, we find that the semantic quality of pseudo-labels is not adequately controlled, resulting in semantic chaos of pseudo-labels. We propose a semantic-stability index to measure the semantic quality of pseudo-labels quantitatively. Based on this index, we propose the Semantic-aware Pseudo-label Reassignment mechanism to generate semantic-like pseudo-labels for all samples. Our approach is model-agnostic and can easily be integrated into existing supervised methods. To demonstrate its generalization ability, we integrate it into two representative algorithms: MAML and EP. The results on three main few-shot benchmarks clearly show that the proposed method achieves significant improvement compared to state-of-the-art models. Notably, our approach also outperforms the corresponding supervised method in three tasks. © 1979-2012 IEEE.

关键词： Latent semantic analysis

来源：评论

学校读者我要写书评

暂无评论

Small object detection based on WavesNET 10

Small object detection based on WavesNET

引用

10th International Conference on Information Systems and Computing Technology, ISCTech 2022

作者： Li, Zhengnan Zhao, Ji Lu, Xiaowei School of Computer and Software Engineering University of Science and Technology Liaoning Anshan China

ISBN: (纸本)9798350332933

In recent years, object detection technology has developed rapidly, especially the FPN has been widely adopted in production application and academic research. Compared with the traditional network, the accuracy of EfficientDet based on the FPN structure has reached 55.1% on the COCO dataset, and the number of parameters and computation cost is also greatly reduced by 9-25 times. However, the EfficientDet training process does not integrate the upper and lower level feature information quickly enough, especially depending on the multi-layer BiFPN overlap and increasing input size and channel number, therefore, we proposed WavesNet based on EfficientDet. First, we integrated seven channels on the basis of BiFPN to make all levels of the fusion more quickly and effectively, and then applied the ComC (Combine Convolution) extraction method in the initial phase of feature extraction. In addition, the EfficientDet fast normalization formula was improved to ensure the numerical stability of the features. Experimental results show the WavesNet has 30% faster convergence rate and 0.5% higher accuracy on COCO dataset than the EfficientDet. The detection effect on small targets is obvious. © 2022 IEEE.

关键词： Object recognition

来源：评论

学校读者我要写书评

暂无评论

Parallel Assembly Sequence Planning Based on Sparrow Search Algorithms 22

Parallel Assembly Sequence Planning Based on Sparrow Search ...

引用

22nd IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2024

作者： He, Haichen Liu, Weibin Zhang, Yinan Song, Shasha Cheng, Yue Liu, Ziheng Xing, Weiwei Beijing Jiaotong University School of Computer Science and Technology China Crrc Information Technology Co. Ltd. China Beijing Jiaotong University School of Software Engineering China

ISBN: (纸本)9798331509712

Assembly sequence planning (ASP) is the most important process in the product lifecycle. This paper aims to propose a parallel assembly sequence planning method based on sparrow search algorithm (SSA) to improve the assembly quality and efficiency. The proposed method divides the problem into assembly unit division (AUD) part and ASP part. In AUD part, we use the markov clustering algorithm (MCA) to solve the AUD problem. In ASP part, the ASP evaluation system is designed for the assembly part sequence planning (APSP) problem and assembly unit sequence planning (AUSP) problem, and the discrete artificial sparrow search algorithm (DASSA) is proposed for solving the discrete APSP problem. The results of comparison experiment verify the effectiveness of the proposed method. Compared with genetic algorithm (GA) and particle swarm optimization (PSO), the proposed DASSA gets the optimal results and fastest convergence. The method proposed solves ASP problem to efficiently perform parallel assembly sequence planning and reduce production time, making it cost-effective in intelligent manufacturing. © 2024 IEEE.

关键词： Particle swarm optimization (PSO)

来源：评论

学校读者我要写书评

暂无评论

SegIt: Empowering Sensor Data Labeling with Enhanced Efficiency and Security 24

SegIt: Empowering Sensor Data Labeling with Enhanced Efficie...

引用

7th International Conference on Machine Learning and Machine Intelligence, MLMI 2024

作者： Zhang, Zhen Abraham, Samuel Lee, Alex Li, Yichen Morota, Gota Ha, Dong Sam Shin, Sook BlacksburgVA United States The Bradley Department of Electrical and Computer Engineering Virginia Tech BlacksburgVA United States School of Computer Science and Software Engineering University of Science and Technology Liaoning Liaoning Anshan China School of Animal Sciences Virginia Tech BlacksburgVA United States

ISBN: (纸本)9798400717833

SegIt is a novel, user-friendly, and highly efficient sensor data labeling tool designed to tackle critical challenges such as data privacy, synchronization accuracy, and memory efficiency inherent in existing labeling tools. While many current sensor data labeling tools provide free online services, they typically necessitate users to upload unlabeled sensor data, alongside video or audio references, to cloud storage for labeling. Nevertheless, such third-party storage exposes user data to potential security risks. SegIt, an innovative open-source tool, provides a software solution for tagging unlabeled sensor data directly on a local computer, ensuring enhanced accuracy, convenience, and, most importantly, data security. © 2024 Copyright held by the owner/author(s).

关键词： Data accuracy

来源：评论

学校读者我要写书评

暂无评论

Consensusless Blockchain: A Promising High-Performance Blockchain Without Consensus 5th

Consensusless Blockchain: A Promising High-Performance Block...

引用

5th International Conference on Blockchain and Trustworthy Systems, BlockSys 2023

作者： Zheng, Jian Wang, Qing Lin, Jianru Huang, Huawei Su, Chunhua Department of Computer Science and Engineering University of Aizu Aizu9650006 Japan School of Computer Science and Engineering Sun Yat-sen University Guangzhou510006 China Humi Technology Co. Ltd. Hainan570000 China School of Software Engineering Sun Yat-sen University Zhuhai519000 China

ISBN: (纸本)9789819981038

Consensus is unnecessary when the truth is available. In this paper, we present a new perspective of rebuilding the blockchain without consensus. When the consensus phase is eliminated from a blockchain, transactions could be canonized quickly using a well-defined universal rule without consuming hashing power. Thus, the transactions per second (TPS) metric of such the consensusless blockchain can be largely boosted. Although consensus blockchain is promising, several technical challenges are also crucial. For example, double-spending attacks and frequent forking events must be prevented, the credit of block’s minting must be carefully defined, and etc. To address those technical challenges, we propose several solutions for our consensusless blockchain (CB), including a naive monotonic scoring mechanism to calculate the ranking of each block in the chain, and a two-stage witness mechanism to add new blocks. The proposed CB chain is promising to offer a simplified and equipment-cheap infrastructure for rich real-world decentralized applications. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Blockchain

来源：评论

学校读者我要写书评

暂无评论

Prediction of Advertising Click-Through Rate Based on Dual Attention Network 4

Prediction of Advertising Click-Through Rate Based on Dual A...

引用

4th International Conference on Machine Learning, Big Data and Business Intelligence, MLBDBI 2022

作者： Zhang, Chao Zhang, Chunna University of Science and Technology LiaoNing School of Computer and Software Engineering Anshan China

ISBN: (纸本)9798350333947

Click-through rate (CTR) prediction refers to whether users will click on the content distributed by manufacturers. In the field of computing advertising and recommendation systems, click-through rate prediction has always been a mainstream hot topic. The relationship between features is often hidden in the data, detailed enumeration will lead to a waste of resources and manpower, and meaningless features will also interfere with and influence the final prediction results. Therefore, the key challenge in CTR task is how to effectively model feature interaction. In this paper, we continue the framework of the traditional deep click rate prediction model, considering the sparsity of the effective feature interaction, we add a feature generation module in the feature input part, and a dual attention network is proposed to further interact and filter the input features. The experiments are conducted on Criteo and Avazu datasets, and the results are improved compared with other models. © 2022 IEEE.

关键词： Marketing

来源：评论

学校读者我要写书评

暂无评论

Breast Microcalcification detection in digital mammograms using Deep Transfer learning approaches 23

Breast Microcalcification detection in digital mammograms us...

引用

9th International Conference on Computing and Data engineering, ICCDE 2023

作者： Rasool, Ehtsham Anwar, Muhammad Junaid Shaker, Bilawal Hashmi, Muhammad Harris Rehman, Khalil Ur Seed, Yousaf Department of Computer Science The University of Alabama Birmingham United States School of Software Engineering Beijing University of Technology Beijing China

ISBN: (纸本)9781450398022

Breast cancer is the most often diagnosed cancer in women affecting one in eight at the age of 80 in US. Breast is the most threatening cancer among women which leads to death. Early diagnosis of breast cancer can save their lives which decreases the mortality rate. Mammography is a standard screening method for breast cancer diagnosis that identifies occurrences of breast cancer in women's at early stages without symptoms. In this study, we employed transfer learning in deep learning to increase the neural network's performance and reduce the false positive rate. In addition, we proposed a pre-trained VGG-19 neural network to extract features of individual microcalcification to predict breast cancer. The proposed method was evaluated on two public databases the CBIS-DDSM and DDSM and achieved 0.98 sensitivities respectively. The proposed method obtained higher sensitivity than other residual neural networks and previous studies. © 2023 ACM.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

Effective Training of PINNs by Combining CMA-ES with Gradient Descent 13

Effective Training of PINNs by Combining CMA-ES with Gradien...

引用

13th IEEE Congress on Evolutionary Computation, CEC 2024

作者： Liu, Lin Yuan, Yuan School of Computer Science and Engineering Beihang University Beijing100191 China State Key Laboratory of Software Development Environment Zhongguancun Laboratory Beijing100191 China

ISBN: (纸本)9798350308365

Physics-Informed Neural Networks (PINNs) have recently received increasing attention, however, optimizing the loss function of PINNs is notoriously difficult, where the landscape of the loss function is often highly non-convex and rugged. Local optimization methods based on gradient information can converge quickly but are prone to being trapped in local minima for training PINNs. Evolutionary algorithms (EAs) are well known for the global search ability, which can help escape from local minima. It has been reported in the literature that EAs show some advantages over gradient-based methods in training PINNs. Inspired by the Memetic Algorithm, we combine global-search based EAs and local-search based batch gradient descent in order to make the best of both word. In addition, since the PINN loss function is composed of multiple terms, balancing these terms is also a challenging issue. Therefore, we also attempt to combine EAs with multiple-gradient descent algorithm for multi-objective optimization. Our experiments provide strong evidence for the superiority of the above algorithms. © 2024 IEEE.

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

computer-Aided Potato Disease Detection by Using Deep Learning Techniques

Computer-Aided Potato Disease Detection by Using Deep Learni...

引用

International Conference on Green Energy, Computing, and Intelligent Technology, GEn-CITy 2023

作者： Razaq, Fareeha Bilal, Muhammad Ramzan, Muhammad Naveed, Muhammad Razzaq, Samreen Department of Computer Science University of Sargodha Sargodha Pakistan Department of Computing and Information Systems School of Engineering and Technology Sunway University Petaling Jaya Malaysia Department of Software Engineering University of Sargodha Sargodha Pakistan School of Electronics and Computer Science University of Southampton Malaysia Iskandar Puteri Malaysia

ISBN: (纸本)9789819998326

Potato is the most widely grown and consumed food throughout the world. There are a number of potato crop diseases that affect production, and these diseases differ in symptoms, circumstances, and controls. Early detection and recognition of disease information can aid in disease prevention and production. This paper presents deep learning models using proposed CNN and pre-trained models for potato disease detection and classification. The proposed model is more efficient and accurate at detection and classification. To perform classification, we used two datasets: PLD and PlantVillage. The Xception CNN model serves as the foundation for our proposed model. It achieved 1.00 accuracy, 0.99 precision, 1.00 recall, and 0.99 F1-score on the PLD dataset. On the PlantVillage dataset, it achieved 1.00 accuracy, 1.00 precision, 1.00 recall, and 1.00 F1-score. We also compared the results of Inception-ResNet-V2, MobileNet-V2, VGG-19, Inception-V3, and Xception models with the performance of our proposed model. For three classes of potato leaves, the proposed CNN model produced more accurate results than other pre-trained models. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Surface Defects Detection of Cylindrical High-Precision Industrial Parts Based on Deep Learning Algorithms: A Review

引用

Operations Research Forum 2024年第3期5卷 58页

作者： Wei, Li Solihin, Mahmud Iwan Saruchi, Sarah ‘Atifah Astuti, Winda Hong, Lim Wei Kit, Ang Chun School of Computer Science and Technology (School of Software) Guangxi University of Science and Technology Guangxi 545006 China Faculty of Engineering Technology and Built Environment UCSI University Kuala Lumpur 56000 Malaysia Faculty of Manufacturing and Mechatronics Engineering Technology Universiti Malaysia Pahang Al-Sultan Abdullah Pahang Pekan 26600 Malaysia Computer Engineering Department Automotive and Robotics Engineering Program BINUS ASO School of Engineering Bina Nusantara University Jakarta Indonesia

High-precision cylindrical parts are critical components across various industries including aerospace, automotive, and manufacturing. Since these parts play a pivotal role in the performance and safety of the systems they are integrated into, they are often subject to stringent quality control measures. Defects on the interior and exterior wall surfaces of these cylindrical parts can severely undermine their function, leading to degraded performance, increased wear, and even catastrophic failures in extreme cases. This article aims to comprehensively summarize the task definition, challenges, mainstream methods, public datasets, evaluation metrics, and other aspects of surface defect detection for high-precision cylindrical parts, in order to help researchers quickly grasp this field. Specifically, the background and characteristics of industrial defect detection are first introduced. Owing to the unique geometric features of cylindrical part surfaces, algorithms and equipment for image data acquisition used in surface defect detection are elaborated in detail. This article presents an extensive overview of state-of-the-art surface defect detection techniques designed for high-precision cylindrical components, all rooted in deep learning. The methods are systematically classified into three main categories: fully supervised, unsupervised, and alternative approaches, based on their data labeling strategies. Additionally, the paper conducts a comprehensive analysis within each category, shedding light on their unique strengths, limitations, and practical use cases. Concluding the discussion, the paper provides insights into future development trends and potential research directions in this field that will lead to manufacturing innovation. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.

关键词： Anomaly detection computer vision Deep learning Defect detection High-precision cylindrical parts Image processing Optical illumination

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：