检索结果-内蒙古大学图书馆

Domain generalization with semi-supervised learning for people-centric activity recognition

science China(Information sciences) 2025年第1期68卷 171-188页

作者： Jing LIU Wei ZHU Di LI Xing HU Liang SONG Academy for Engineering & Technology Fudan University Shanghai East-bund Research Institute on Networking Systems of AI School of Optoelectronic Information and Computer Engineering University of Shanghai for Science & Technology

People-centric activity recognition is one of the most critical technologies in a wide range of real-world applications,including intelligent transportation systems, healthcare services, and brain-computer interfaces. Large-scale data collection and annotation make the application of machine learning algorithms prohibitively expensive when adapting to new tasks. One way of circumventing this limitation is to train the model in a semi-supervised learning manner that utilizes a percentage of unlabeled data to reduce the labeling burden in prediction tasks. Despite their appeal, these models often assume that labeled and unlabeled data come from similar distributions, which leads to the domain shift problem caused by the presence of distribution gaps. To address these limitations, we propose herein a novel method for people-centric activity recognition,called domain generalization with semi-supervised learning(DGSSL), that effectively enhances the representation learning and domain alignment capabilities of a model. We first design a new autoregressive discriminator for adversarial training between unlabeled and labeled source domains, extracting domain-specific features to reduce the distribution gaps. Second, we introduce two reconstruction tasks to capture the task-specific features to avoid losing information related to representation learning while maintaining task-specific consistency. Finally, benefiting from the collaborative optimization of these two tasks, the model can accurately predict both the domain and category labels of the source domains for the classification task. We conduct extensive experiments on three real-world sensing datasets. The experimental results show that DGSSL surpasses the three state-of-the-art methods with better performance and generalization.

关键词： activity recognition deep learning domain generalization semi-supervised learning adversarial training

来源：评论

学校读者我要写书评

暂无评论

A lightweight approach for intrusion detection in WSNs based on DCGAN

引用

International Journal of Information Technology (Singapore) 2025年第2期17卷 951-957页

作者： Devi, Manu Nandal, Priyanka Sehrawat, Harkesh Department of Computer Science and Engineering University Institute of Engineering and Technology Maharshi Dayanand University Haryana India Department of Computer Science and Engineering Maharaja Surajmal Institute of Technology New Delhi India

Ensuring strong security procedures is crucial in the rapidly advancing realm of wireless sensor networks (WSNs) in order to protect sensitive data and preserve network integrity. The resource limitations and unpredictable environments that characterize WSNs frequently provide challenges for traditional intrusion detection systems (IDS). To overcome these issues, this research introduces a unique lightweight intrusion detection system that utilizes Deep Convolutional Generative Adversarial Networks (DCGAN). The proposed system employs wireless sensor network blackhole flooding selective forwarding (WSN-BFSF) dataset for training and evaluation. The suggested approach improves the IDS’s detection capabilities in addition to reducing computing overhead. The findings show that the lightweight DCGAN-based IDS provides an improved defense against new security threats, making it a good fit for real-time deployment in WSNs. Here, the performance metrics like receiver operating characteristic (ROC) is evaluated to check the efficiency of the model. The proposed model is assessed with the existing studies based on metrics such as accuracy, precision, recall and f1-score. The results show that the proposed approach attains the highest accuracy (94%) thus outperforming the existing methods and studies. © Bharati Vidyapeeth's institute of computer Applications and Management 2024.

关键词： Deep Learning (DL) Generative Adversarial Networks (GAN’s) Intrusion Detection System (IDS) Lightweight DCGAN Wireless Sensor Networks (WSN’s)

来源：评论

学校读者我要写书评

暂无评论

BGRU-MTRA: bilinear GRU networks with multi-path temporal residual attention for suspicious activity recognition

引用

Neural Computing and Applications 2025年第1期37卷 185-212页

作者： Pandey, Ajeet Kumar, Piyush Computer Science and Engineering National Institute of Technology Patna Ashok Rajpath Bihar Patna800005 India

Suspicious activity recognition (SAR) is an active research field in computer vision and image processing due to the rapid demand for intelligent video surveillance systems. However, current automated systems focus too much on the temporal dynamics of events in videos and ignore the importance of spatial dynamics. The existing methods for SAR are complex and demand substantial resources. Hence, this paper proposes a novel trade-off architecture for SAR integrating the strengths of the multi-path temporal residual attention (MTRA) module with the bilinear-gated recurrent unit (BGRU) module. The MTRA module combines spatial feature extraction (SFE) with temporal residual attention network (TRAN) blocks for improving SAR by enhancing resilience to variations in object sizes, viewpoints, and motion patterns. It selects relevant action features, addresses vanishing gradient issues, and reduces spatial dimension. The BGRU module preserves spatial features, improving the model’s ability to recognize fine-grained features and complex temporal patterns for effectively recognizing suspicious activities. The BGRU-MTRA system yields recognition accuracy of 93.28%, 98.97%, 99.86%, and 99.66%, and 48.42% on the benchmark hybrid crime action (HCA), UT interaction (UTI), CAVIER, hockey fight (HF), and the most challenging UCF-crime datasets, respectively. Through comprehensive experiments, it is clear that the proposed method outperforms state-of-the-art methods over benchmark datasets with reduced parameters. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Spatio-temporal data

来源：评论

学校读者我要写书评

暂无评论

Dynamic-EC:an efficient dynamic erasure coding method for permissioned blockchain systems

引用

Frontiers of computer science 2025年第1期19卷 15-28页

作者： Mizhipeng ZHANG Chentao WU Jie LI Minyi GUO Department of Computer Science and Engineering Shanghai Jiao Tong UniversityShanghai 200200China Yancheng Blockchain Research Institute Hengyang 421200China

Blockchain as a decentralized storage technology is widely used in many *** has extremely strict requirements for reliability because there are many potentially malicious ***,blockchain is a chain storage structure formed by interconnecting blocks1),which are stored by full replication method,where each node stores a replica of all blocks and the data consistency is maintained by the consensus *** decrease the storage overhead,previous approaches such as BFT-Store and Partition Chain store blocks via erasure ***,existing erasure coding based methods utilize static encoding schema to tolerant f malicious nodes,but in the typical cases,the number of malicious nodes is much smaller than f as described in previous *** redundant parities to tolerate excessive malicious nodes introduces unnecessary storage *** solve the above problem,we propose Dynamic-EC,which is a Dynamic Erasure Coding method in permissioned blockchain *** key idea of Dynamic-EC is to reduce the storage overhead by dynamically adjusting the total number of parities according to the risk level of the whole system,which is determined by the number of perceived malicious nodes,while ensuring the system *** demonstrate the effectiveness of Dynamic-EC,we conduct several experiments on an open source blockchain software *** results show that,compared to the state-of-the-art erasure coding methods,Dynamic-EC reduces the storage overhead by up to 42%,and decreases the average write latency of blocks by up to 25%,respectively.

关键词： blockchain Byzantine Fault Tolerance(BFT) erasure coding consensus reputation evaluation

来源：评论

学校读者我要写书评

暂无评论

REGAC: Multi-Class Offensive Content Identification Using Graphical Approach

IAENG International Journal of Computer Science

引用

IAENG International Journal of computer science 2025年第4期52卷 931-952页

作者： Chinivar, Sneha Roopa, M.S. Arunalatha, J.S. Venugopal, K.R. Department of Computer Science and Engineering Bangalore University K R Circle Karnataka Bengaluru India Department of Computer Science and Engineering Nitte Meenakshi Institute of Technology Karnataka Bengaluru India Department of Computer Science and Engineering Bangalore University K R Circle Karnataka Bengaluru India Karnataka Bengaluru India

Online offensive behaviour continues to rise with the increasing popularity and use of social media. Various techniques have been used to address this issue. However, most existing studies consider offensive content identification as a binary or ternary problem, disregarding the potential for multi-class classification of derogatory content. In this work, we propose RoBERTa Embedding-based Graphical Approach Classifier (REGAC), which aims to designate the identified offensive content into varied classes,like age, gender, ethnicity, religion, and others to understand what exact qualities the bully usually targets in their victim so that pertinent measures can be taken to battle them. Additionally, the models’ efficiency is evaluated using an unbalanced dataset to identify types of offensive content. The word embeddings are generated for a balanced dataset using RoBERTa embedding model to bring out the best-fitting vectors. These vectors are then input into traditional Machine Learning algorithms (SVM, KNN, Logistic Regression, Random Forest, and XGBoost) and graph-based algorithms (Graph Convolution Network (GCN), GraphSAGE, and Graph Attention Network (GAT)) for fine-grained categorization of offensive textual content. The experimental results demonstrate the efficiency of combating social media’s offensive content of the proposed work with higher Precision, Recall, F1-Score, and Accuracy compared with most state-of-art approaches. © (2025.0), (International Association of Engineers). All Rights Reserved.

关键词： Graph embeddings

来源：评论

学校读者我要写书评

暂无评论

Halo reduction multi-exposure image fusion technique

引用

Multimedia Tools and Applications 2025年第13期84卷 12347-12370页

作者： Sharif, Rizwan Amin, Benish Sukhia, Komal Nain Department of Electrical Engineering Institute of Space Technology Islamabad Pakistan Department of Computer Science Institute of Space Technology Islamabad Pakistan

Multi-exposure image fusion (MEF) involves combining images captured at different exposure levels to create a single, well-exposed fused image. MEF has a wide range of applications, including low light, low contrast, night photography, medical imaging, and remote sensing. However, MEF methods often face issues like artifacts, halos around edges, color inconsistencies, noise amplification, and difficulty in preserving fine details. Moreover, assessing the quality of fused images objectively is complex due to the subjective nature of human perception. Solving the challenges is essential to developing efficient MEF techniques that produce high-quality results across various scenarios. The proposed technique introduces an approach to handling halo artifacts and implementing MEF. The Dense Scale-Invariant Feature Transform (DSIFT) is used to capture vital information about image brightness, texture, and edges from source images. Three weight maps are computed from the local mean, signal strength, and the global gradient for initial weight estimation. The local mean represents the brightness of specific image areas, signal strength preserves essential details like textures and edges while reducing image noise, and global gradient helps identify regions with significant pixel value shifts across multiple exposure images. The weight maps are then combined using a weighted average and a Gaussian smoothing filter is applied to reduce inherent noise and discontinuities in the original weights. Subsequently, pyramid decomposition is performed to generate a fused image. The efficiency of the proposed method is extensively tested on challenging multi-exposure image sequences. The results of the proposed approach demonstrate its superiority in both subjective evaluation and objective metrics like the MEF-Structural Similarity Index (MEF-SSIM), Natural Image Quality Evaluator (NIQE) and Gradient based performance measure (QAB/f). © The Author(s), under exclusive licence to Springer

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

Function Scheduling with Data Security in Serverless Computing Systems 17

Function Scheduling with Data Security in Serverless Computi...

引用

17th International Conference on COMmunication Systems and NETworkS, COMSNETS 2025

作者： Saha, Srestha Pandey, Arvind Addya, Sourav Kanti Brata Nath, Shubha Indian Institute of Information Technology Department of Computer Science and Engineering Guwahati India SRM Institute of Science and Technology Department of Computer Science and Engineering Tamil Nadu India National Institute of Technology Karnataka Department of Computer Science and Engineering Surathkal India

ISBN: (纸本)9798331531195

In serverless computing, the service provider takes full responsibility for function management. However, serverless computing has many challenges regarding data security and function scheduling. To address these challenges, we have proposed a system to secure the data of an end-user. We also aim to meet the quality of service (QoS) for the end-user requests. This work presents a Simulated Annealing-based optimization algorithm for function placement. Also, we have Hyperledger Fabric, a blockchain framework in the system architecture for securing the data of an end-user. We have conducted experiments in Amazon Elastic Compute Cloud (EC2) taking virtual machine instances. The experiments in Amazon EC2 indicate that the proposed system secures the data and enhances the end-user's QoS. © 2025 IEEE.

关键词： Simulated annealing

来源：评论

学校读者我要写书评

暂无评论

Automated face recognition using deep learning technique and center symmetric multivariant local binary pattern

引用

Neural Computing and Applications 2025年第1期37卷 263-281页

作者： Sekhar, J.C. Josephson, P. Joel Chinnasamy, A. Maheswari, M. Sankar, S. Kalangi, Ruth Ramya Department of Computer Science and Engineering NRI Institute of Technology Andhra Pradesh Guntur India Department of Electronics and Communication Engineering Malla Reddy Engineering College Telangana Hyderabad India Department of Data Science and Business Systems School of Computing SRMIST Kattankulathur Chennai India Department of Computer Science and Engineering Panimalar Engineering College Chennai India Department of Computer Science and Engineering Saveetha School of Engineering SIMATS Chennai India Department of Computer Science and Engineering Koneru Lakshmaiah Education Foundation Andhra Pradesh Vijayawada India

Researchers have recently created several deep learning strategies for various tasks, and facial recognition has made remarkable progress in employing these techniques. Face recognition is a noncontact, nonobligatory, acceptable, and harmonious biometric recognition method with a promising national and social security future. The purpose of this paper is to improve the existing face recognition algorithm, investigate extensive data-driven face recognition methods, and propose a unique automated face recognition methodology based on generative adversarial networks (GANs) and the center symmetric multivariable local binary pattern (CS-MLBP). To begin, this paper employs the center symmetric multivariant local binary pattern (CS-MLBP) algorithm to extract the texture features of the face, addressing the issue that C2DPCA (column-based two-dimensional principle component analysis) does an excellent job of removing the global characteristics of the face but struggles to process the local features of the face under large samples. The extracted texture features are combined with the international features retrieved using C2DPCA to generate a multifeatured face. The proposed method, GAN-CS-MLBP, syndicates the power of GAN with the robustness of CS-MLBP, resulting in an accurate and efficient face recognition system. Deep learning algorithms, mainly neural networks, automatically extract discriminative properties from facial images. The learned features capture low-level information and high-level meanings, permitting the model to distinguish among dissimilar persons more successfully. To assess the proposed technique’s GAN-CS-MLBP performance, extensive experiments are performed on benchmark face recognition datasets such as LFW, YTF, and CASIA-WebFace. Giving to the findings, our method exceeds state-of-the-art facial recognition systems in terms of recognition accuracy and resilience. The proposed automatic face recognition system GAN-CS-MLBP provides a solid basis for a

关键词： Principal component analysis

来源：评论

学校读者我要写书评

暂无评论

Adaptive Attention Mechanisms Based Multicriteria Recommender Systems for Improved User Preference Discovery

引用

SN computer science 2025年第4期6卷 1-16页

作者： Rajput, Ishwari Singh Tewari, Anand Shanker Tiwari, Arvind Kumar Department of Computer Science & Engineering National Institute of Technology Patna Bihar Patna India Department of Computer Science & Engineering Kamla Nehru Institute of Technology Sultanpur (UP) India

Personalized recommender systems are becoming more popular to reduce the issue of information overload. It is also observed that the recommendations provided by multi-criteria recommender system (MCRS) are more accurate and precise in comparison to conventional single-criteria recommender systems. MCRS incorporates customers’ preferences based on multiple criteria while recommending any item. Recent research has shown that MCRS can efficiently use implicit and explicit input across various dimensions to make exact recommendations across multiple domains. Due to use of multiple criteria, existing MCRS methods often struggle with effectively aggregating and interpreting these criteria, making it hard to accurately determine user preferences. In this article, we introduce a novel adaptive attention mechanism that dynamically assigns weights to each criterion based on user preferences. Each criterion is initially weighted based on its importance to a user. An attention mechanism is used to understand the exact user’s preference for each criterion. Later, the nearest-neighbor method finds users with similar interest to predict unknown ratings which is further used for Top-N recommendations. The performance of the proposed approach (AM-MCRS) is validated on Yahoo! movies multi-criteria dataset having 1716 unique users and 965 different movies and a comparative analysis is performed with existing single and multi-criteria recommender systems. The results clearly show that the AM-MCRS methodology outperform baseline methods, achieving the highest precision (0.9412), F-score (0.9343), accuracy (0.929), and coverage (0.9407), while also observing a significant reduction in MAE (0.7623) and RMSE (0.9145). These findings highlight the capability of AM-MCRS to improve recommendation accuracy and dependability across various domains. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Attention mechanism Deep learning Dynamic user preference Multi-criteria recommender system Nearest-neighbor

来源：评论

学校读者我要写书评

暂无评论

HViTML: Hybrid vision transformer with machine learning-based classification model for glaucomatous eye

引用

Multimedia Tools and Applications 2025年 1-24页

作者： Singh, Piyush Bhushan Singh, Pawan Dev, Harsh Batra, Devanshu Chaurasia, Brijesh Kumar Department of Information Technology Pranveer Singh Institute of Technology Uttar Pradesh Kanpur India Department of Computer Science and Engineering Amity School of Engineering and Technology Amity University Uttar Pradesh Lucknow India Department of Computer Science and Engineering Babu Banarasi Das University Uttar Pradesh Lucknow India Department of Computer Science and Engineering Pranveer Singh Institute of Technology Uttar Pradesh Kanpur India

Glaucoma is currently one of the most significant causes of permanent blindness. Fundus imaging is the most popular glaucoma screening method because of the compromises it has to make in terms of portability, size, and cost. In recent years, convolution neural networks (CNNs) have revolutionized computer vision. Convolution is a "local" CNN technique that is only applicable to a small region surrounding an image. Vision Transformers (ViT) use self-attention, which is a "global" activity since it collects information from the entire image. As a result, the ViT can successfully gather distant semantic relevance from an image. This study examined several optimizers, including Adamax, SGD, RMSprop, Adadelta, Adafactor, Nadam, and Adagrad. With 1750 Healthy and Glaucoma images in the IEEE fundus image dataset and 4800 healthy and glaucoma images in the LAG fundus image dataset, we trained and tested the ViT model on these datasets. Additionally, the datasets underwent image scaling, auto-rotation, and auto-contrast adjustment via adaptive equalization during preprocessing. The results demonstrated that preparing the provided dataset with various optimizers improved accuracy and other performance metrics. Additionally, according to the results, the Nadam Optimizer improved accuracy in the adaptive equalized preprocessing of the IEEE dataset by up to 97.8% and in the adaptive equalized preprocessing of the LAG dataset by up to 92%, both of which were followed by auto rotation and image resizing processes. In addition to integrating our vision transformer model with the shift tokenization model, we also combined ViT with a hybrid model that consisted of six different models, including SVM, Gaussian NB, Bernoulli NB, Decision Tree, KNN, and Random Forest, based on which optimizer was the most successful for each dataset. Empirical results show that the SVM Model worked well and improved accuracy by up to 93% with precision of up to 94% in the adaptive equalization preprocess

关键词： Decision trees

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：