检索结果-内蒙古大学图书馆

Energy-Efficient Distributed Learning for NOMA-Based Unmanned Aerial Agent-assisted MEC Networks

IEEE Internet of Things Journal 2025年

作者： Consul, Prakhar Budhiraja, Ishan Garg, Deepak Kumar, Neeraj Rodrigues, Joel J. P. C. Almuhaideb, Abdullah IILM University School of Computer Science and Engineering Department of Computing and Security Greater Noida India Bennett University School of Computer Science Engineering and Technology Greater Noida India SR University School of Artificial Intelligence and Computer Science Telangana Warangal India Thapar Institute of Engineering and Technology Department of Computer Science and Engineering Punjab Patiala India Imam Abdulrahman Bin Faisal University Department of Networks and Communications College of Computer Science and Information Technology Saudi Arabia PI Teresina Brazil

The Internet of Things (IoT) has become a revolutionary concept that connects various devices and systems to enable smooth communication and data exchange. In this vast network, unmanned aerial agents (UAA)-assisted mobile edge computing (MEC) communication plays a crucial role in facilitating direct interaction between edge devices. This aspect of IoT goes beyond traditional interactions between humans and machines. It creates a dynamic environment where devices collaborate autonomously, share information, and perform tasks. UAA-assisted MEC network offers several benefits such as supports short range communication, reduced delay, improved scalability, and enhanced energy efficiency. Furthermore, for the purpose of enhancing the widespread interconnection and exceptionally dependable minimal delay in the fifth generation (5G) and beyond network, the utilization of non-orthogonal multiple access (NOMA) can be considered. Within this context, the impact of federated learning (FL) on NOMA-based UAV-assisted MEC network in wirelesspowered communication networks is examined. Initially, the transmitters extract energy from the radio frequency signals emitted by the MEC server. Subsequently, the transmitters utilize NOMA to establish communication with the receivers by utilizing the stored harvested energy. The formulation of a stochastic optimization problem is proposed with the aim of improving energy consumption (EC) and minimizing delay. Results indicate that the proposed scheme exhibit superior accuracy compared to baseline schemes, achieving an accuracy 98.37% after 59 communication rounds. The FL is employed to attain the objective and accelerate the local training data across the UAA-assisted MEC network. © 2025 IEEE.

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

Neural Networks Meet Neural Activity: Utilizing EEG for Mental Workload Estimation 27th

Neural Networks Meet Neural Activity: Utilizing EEG for Ment...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Siddhad, Gourav Roy, Partha Pratim Kim, Byung-Gyu Department of Computer Science and Engineering Indian Institute of Technology Roorkee Uttarakhand Roorkee247667 India Division of AI Engineering Sookmyung Women’s University Seoul04310 Korea Republic of

ISBN: (纸本)9783031781940

Electroencephalography (EEG) offers non-invasive, real-time mental workload assessment, which is crucial in high-stakes domains like aviation and medicine and for advancing brain-computer interface (BCI) technologies. This study introduces a customized ConvNeXt architecture, a powerful convolutional neural network, specifically adapted for EEG analysis. ConvNeXt addresses traditional EEG challenges like high dimensionality, noise, and variability, enhancing the precision of mental workload classification. Using the STEW dataset, the proposed ConvNeXt model is evaluated alongside SVM, EEGNet, and TSception on binary (No vs SIMKAP task) and ternary (SIMKAP multitask) class mental workload tasks. Results demonstrated that ConvNeXt significantly outperformed the other models, achieving accuracies of 95.76% for binary and 95.11% for multi-class classification. This demonstrates ConvNeXt’s resilience and efficiency for EEG data analysis, establishing new standards for mental workload evaluation. These findings represent a considerable advancement in EEG-based mental workload estimation, laying the foundation for future improvements in cognitive state measurements. This has broad implications for safety, efficiency, and user experience across various scenarios. Integrating powerful neural networks such as ConvNeXt is a critical step forward in non-invasive cognitive monitoring. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

Virtual viewpoint reconstruction utilizing monocular images with complex backgrounds 4

Virtual viewpoint reconstruction utilizing monocular images ...

引用

4th International Computational Imaging Conference, CITA 2024

作者： Li, Ling-Ling Wang, Tianchen Zhang, Huiling Zhao, Jian Xia, Jun School of Computer Engineering Nanjing Institute of technology No.1 Hongjing Road Jiangsu Nanjing211167 China School of Electronic Science and Engineering Southeast University Nanjing210096 China

ISBN: (纸本)9781510688834

In recent years, significant advancements have been made in the field of three-dimensional light field display, positioning it as one of the most promising display technologies. To enhance the effectiveness and realism of 3D information representation, a dense sequence of virtual viewpoints is required for optimal display. However, the acquisition of high-quality virtual viewpoint sequences poses one of the most significant challenges in this technology. Moreover, when relying solely on a single 2D image as reference information, reconstruction of virtual viewpoints often encounters issues such as ghosting artifacts, contour distortions, blurring effects, and gradient inconsistencies. Therefore, a virtual viewpoint generation algorithm based on large-scale neural networks and local boundary restoration is presented. The target view depth map is generated with precise and distinct boundary information based on large model neural networks. Additionally, an efficient local boundary matching algorithm effectively fills occlusion areas. This approach enables the generation of high-quality dense virtual view sequences even in complex backgrounds, offering significant advantages. © 2025 SPIE.

关键词： Virtual environments

来源：评论

学校读者我要写书评

暂无评论

WebShield 5.0: Harnessing AI and NLP to combat web threats in Industry 5.0

引用

Alexandria engineering Journal 2025年 127卷 677-689页

作者： Verma, Priyanka O'Shea, Donna Newe, Thomas Vidyarthi, Ankit Gupta, Deepak Ali, Jabir Aldawsari, Hamad Breslin, John G. School of Computer Science University of Galway University road GalwayH91 TK33 Ireland Department of Electronics and Computer Engineering University of Limerick Castletroy LimerickV94 T9PX Ireland Digital Engineering University of Limerick Castletroy LimerickV94 T9PX Ireland Department of CSE&IT Jaypee Institute of Information Technology sector 62 Noida201309 India Maharaja Agrasen Institute of Technology Delhi India School of Computer Science Engineering and Technology Bennett University Greater Noida Delhi India Department of Computer Science Haql University College University of Tabuk Tabuk Saudi Arabia Data Science Institute University of Galway University road GalwayH91 TK33 Ireland

Industry 5.0 characterized by the integration of human intelligence and advanced technologies is inherently more connected and interdependent than previous industrial paradigms. This increased connectivity exposes to various web-based attacks and calls for strong security controls. To address the challenges and enhance attack detection, this paper introduces Ingress Manager (IM), a novel approach that amalgamates Natural Language Processing (NLP) with Machine Learning (ML) to mitigate web-based threats. By combining multimodal data and utilizing the Mayfly optimization algorithm for feature selection, IM carries out a thorough analysis for efficient web-based attack detection. Mayfly Optimization is considered to be a variation of Particle Swarm Optimization (PSO), combining the benefits of Firefly Algorithm, Genetic Algorithm (GA), and PSO. Experiments on the HTTP CSIC-2010 dataset show that this integration is effective, as evidenced by the notable gains in accuracy, precision, and F-score above baseline models. Notable performance metrics such as accuracy of 98.5753% along with thorough component analysis (ablation study) add deeper understanding to the proposed approach. The paper's contributions lie in its utilization of Industry 5.0 principles, the incorporation of Mayfly optimization for feature selection, and the innovative combination of NLP and ML for robust web-based attack detection. © 2025

关键词： Genetic algorithms

来源：评论

学校读者我要写书评

暂无评论

Artificial neural networks and noncontact microwave NDT for evaluation of polypropylene fiber concrete

引用

Asian Journal of Civil engineering 2025年第1期26卷 273-292页

作者： Nimer, Hamsa Ismail, Rabah Al-Mattarneh, Hashem Khodier, Mohanad Jaradat, Yaser Rawashdeh, Adnan Rawashdeh, Mohammad Civil Engineering Department Yarmouk University Irbid Jordan Civil Engineering Department Jadara University Irbid Jordan Computer Science & amp Information Technology Yarmouk University Irbid Jordan Communication Engineering Department Yarmouk University Irbid Jordan

Polypropylene fibers are extensively incorporated into reinforced concrete to enhance performance aspects such as crack resistance, flexural and tensile strength, fire resistance, and overall durability. However, current methods for evaluating factors like fiber inclusion percentage, distribution, and orientation within the concrete matrix are often limited, destructive, and time-consuming. This study explores developing and applying a non-contact microwave non-destructive method (NMNDT) for assessing polypropylene fiber-reinforced concrete. The NMNDT system measures the reflection and transmission characteristics of microwave signals through the concrete, correlating these properties with the physical and mechanical characteristics of the material. Key findings indicate a strong correlation between microwaves’ reflection and transmission properties, the quality of fiber distribution, and the fiber content within the concrete. For instance, the study found that the reflection coefficient (S₁₁) increased from 0.36 to 0.39, with fiber content varying from 0.5 to 1.5 kg/m³, while the transmission coefficient (S₂₁) decreased from 0.46 to 0.38 over the same range. The compressive strength of fiber-reinforced concrete was predicted with a correlation coefficient (R) of 0.98 using artificial neural networks (ANN). These microwave properties can predict mechanical properties such as tensile and compressive strength, with the ANN model achieving more than 97% accuracy. The study highlights the innovative potential of microwave technology as a non-invasive evaluation technique for polypropylene fiber-reinforced concrete, offering a promising avenue for rapid and non-destructive quality control and performance assessment. The integration of ANN further enhances the predictability of the strength properties of polypropylene fiber-reinforced concrete, significantly contributing to advancements in the field of fiber-reinforced concrete evaluation and quality

关键词： Artificial neural network (ANN) Concrete technology Mechanical properties Nondestructive Polypropylene fibers concrete

来源：评论

学校读者我要写书评

暂无评论

Enhanced DNA Sequence Classification Through Hyperparameter Optimized Convolutional Neural Networks

Enhanced DNA Sequence Classification Through Hyperparameter ...

引用

International Conference on Electrical, computer and Communication engineering (ECCE)

作者： Md Mehedi Hasan Hura Jannat Poly Rayhanuzzaman Rayhanuzzaman Azmain Yakin Srizon Department of Computer Science and Engineering Port City International University Chattogram Bangladesh Department of Computer Science and Engineering International Islamic University Chittagong Chattogram Bangladesh Department of Computer Science & Engineering Rajshahi University of Engineering & Technology Rajshahi Bangladesh

ISBN: (数字)9798350357509

ISBN: (纸本)9798350357516

DNA sequence classification is a fundamental task in bioinformatics, with applications ranging from gene prediction to disease diagnosis. Convolutional Neural Networks (CNNs) are a deep learning-based method that is suggested for DNA sequence classification. Two significant biological datasets, Promoter and Splice, are utilized for evaluation. The raw DNA sequences are transformed into numerical representations through k-mer feature extraction, serving as input for the CNN model. Multiple convolutional and pooling layers, followed by a fully linked layer for classification, make up the carefully planned CNN architecture. Extensive hyperparameter tuning is conducted to optimize the model, including adjustments to parameters such as learning rate, kernel size, and number of layers. Experimental results indicate that the tuned CNN model achieves a classification accuracy of 97% on both the Promoter and Splice datasets. The performance of the proposed model is compared with other baseline machine learning models, demonstrating its superior effectiveness. This study underscores the utility of CNNs in DNA sequence classification tasks and provides valuable insights for genomic research and applications in bioinformatics.

关键词： Learning systems Biological system modeling Machine learning Feature extraction Numerical models Convolutional neural networks Medical diagnosis Bioinformatics Kernel Tuning

来源：评论

学校读者我要写书评

暂无评论

Near-Field Localization with Antenna Arrays in the Presence of Direction-Dependent Mutual Coupling

引用

IEEE Transactions on Vehicular Technology 2025年第5期74卷 7033-7048页

作者： Ebadi, Zohreh Molaei, Amir Masoud Alexandropoulos, George C. Abbasi, Muhammad Ali Babar Cotton, Simon Tukmanov, Anvar Yurduseven, Okan Queen's University Belfast Centre for Wireless Innovation School of Electronics Electrical Engineering and Computer Science BelfastBT3 9DT United Kingdom National and Kapodistrian University of Athens Department of Informatics and Telecommunications Athens15784 Greece University of Illinois Chicago Department of Electrical and Computer Engineering Department ChicagoIL60601 United States BT Group BT Labs Adastral Park Ipswich United Kingdom

Localizing near-field sources considering practical arrays is a recent challenging topic for next generation wireless communication systems. Practical antenna array apertures with closely spaced elements exhibit direction-dependent mutual coupling (MC), which can significantly degrade the performance localization techniques. A conventional method for near-field localization in the presence of MC is the three-dimensional (3D) multiple signal classification (MUSIC) technique, which, however, suffers from extremely high computational complexity. Recently, two-dimensional (2D) search alternatives have been presented, exhibiting increased complexity still for direction-dependent MC scenarios. In this paper, we devise a low complexity one-dimensional (1D) iterative method based on an oblique projection operator (IMOP) that estimates direction-dependent MC and the locations of multiple near-field sources. The proposed method first estimates the initial direction of arrival (DOA) using an approximate wavefront model, and then, computes the initial MC and range of the near-field source using the exact wavefront model. Afterwards, at each iteration, the oblique projection operator is used to isolate components associated with one source from those of other sources. The DOA and range of this one source are estimated using the exact wavefront model and 1D searches. Finally, the direction-dependent MC is estimated for each pair of the estimated DOA and range. The performance of the proposed near-field localization approach is comprehensively investigated and verified using both a full-wave electromagnetic solver and synthetic simulations. It is showcased that our IMOP scheme performs almost similarly to a state-of-the-art approach but with a 42 times less computational complexity. © 2025 IEEE.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Detection of Medicinal Plants Using Machine Learning 12th

Detection of Medicinal Plants Using Machine Learning

引用

12th International Conference on Recent Trends in Computing, ICRTC 2024

作者： Ravikumar, S. Eugene Berna, I. Babu, R. Arockia Raj, Y. Vijay, K. Department of Computer Science and Engineering Vel Tech Rangarajan Dr Sagunthala R&D Institute of Science and Technology Chennai India Department of Artificial Intelligence and Machine Learning Bannari Amman Institute of Technology Erode India Department of Computational Intelligence School of Computing SRM Institute of Science and Technology Kattankulathur Campus Chennai India Department of Computer Science and Engineering PSNA College of Engineering and Technology Dindigul India Department of Artificial Intelligence and Machine Learning Rajalakshmi Engineering College Chennai India

ISBN: (纸本)9789819789450

An exciting new direction in plant categorization and identification is the use of machine learning techniques for medicinal plant detection. Using a variety of photo datasets containing medicinal plants, this technique trains models using ANNs. The ML model learns the complex traits that distinguish various plant species by using data from pre-trained models or by developing bespoke architectures. Thanks to the trained model’s precision in visually recognizing therapeutic plants, everyone from researchers to ecologists can reap the benefits. Methods for medicinal plant identification that are both accurate and objective are crucial to biodiversity research and attempts to safeguard resources. Considerations of ethics, model calibration, and data curation should be incorporated into the construction of such systems. Gather any and all images of plants used for medical purposes. Various viewpoints, lighting scenarios, and plant types will enhance the set. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Plant diseases

来源：评论

学校读者我要写书评

暂无评论

Energy-Aware Framework for Assessing Cryptographic Algorithms in Wireless Sensor Networks

引用

SN computer science 2025年第4期6卷 1-17页

作者： Pooja Shalu Saini, Neha Baghela, Vishwadeepak Singh Khan, Mohd Anas School of Computer Science Engineering and Technology Bennett University Uttar Pradesh Greater Noida India Manav Rachna University Faridabad Haryana Faridabad India Government College Chhachhrauli Haryana Yamuna Nagar India SCSE Galgotias University Uttar Pradesh Greater Noida India Department of Computer Engineering Jamia Millia Islamia Delhi India

Wireless sensor networks have found applications in several areas such as agriculture, healthcare, weather forecasting, and the Internet of Things. Due to their extensive use WSNs are susceptible to several significant network threats. Therefore, they necessitate the implementation of advanced security measures. An issue of significant concern is the possible violation of user privacy resulting from the transmission of data. The rapid increase in online transactions and the vast volume of data generated on a daily basis have elevated information security to an important concern in the present era of the internet. Cryptography serves as a durable remedy to these issues by bolstering the confidentiality of data transmitted across the internet. This study analysed many characteristics, such as the total execution time, memory needs, number of packets sent to the base station, number of dead nodes, remaining energy after transmission, and the use of RSA algorithm and Elliptic Curve Cryptography. The simulation is executed in MATLAB utilising pre-defined parameters and constants. A WSN observation scenario was demonstrated with a fixed number of rounds (200) and varying numbers of nodes (5, 10, 15, and 20). The researchers computed the average ratio of packets transferred to the base station, inoperative nodes, and leftover energy after transmission. The upper limit for the number of packets that can be successfully transmitted to the base station using AES is 55%. This is subject to the condition that at least 40% of the nodes are inactive and a maximum of 45% of the nodes have residual energy. The primary achievement of the current work is the identification of the most effective energy-efficient encryption algorithm, as previous research has not focused extensively on this particular characteristic. Based on the simulation results, AES outperforms RSA and Elliptic Curve Cryptography in terms of packet transmission, inactive nodes, and energy conservation. © The Author

关键词： AES Asymmetric algorithm Cryptography ECC Energy left RSA Symmetric algorithms

来源：评论

学校读者我要写书评

暂无评论

UAV -Based Object Tracking Using Siamese Networks and Kalman Filtering

UAV -Based Object Tracking Using Siamese Networks and Kalman...

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Ramanan, S.V. Rajasekar, M. Shanmugapriya, G. Sujitha, R. Akshya, J. Choudhry, Mani Deepak P Pg Institute of Technology Department of Electronics and Communication Engineering Coimbatore India Vsb college of Engineering Department of Computer Science and Engineering Technical Campus Coimbatore India Dr.N G.P Institute of Technology Department of Computer Science and Business Systems Coimbatore India Christ The King Engineering College Department of Information Technology Coimbatore India Srm Institute of Science and Technology Department of Computational Intelligence Chennai India Srm Institute of Science and Technology Department of Computing Technologies Chennai India

ISBN: (纸本)9798331509828

The ability to match features and keep track of objects in changing dynamic environments is still an important problem, particularly due to varying noise levels, diverse datasets, and high dimensional feature spaces. Euclidean Distance and Cosine Similarity based matching are existing methods that commonly do not generalize well, particularly in the case of cross domain scenarios, where noise and motion dynamics are powerful factors to limit the accuracy of tracking. However, Deep Metric Learning approaches can extract robust features at the cost of computational efficiency and are unsuitable for real-time applications. This research proposes a Siamese Network framework to solve these problems by using robust feature matching and multi-object tracking. Based on that, the proposed framework uses a dual-branch network architecture in which feature embeddings extracted from paired inputs are mapped through the same shared latent space. Furthermore, the adaptive learning rate leads to efficient convergence during the training. Experimental results show that the Siamese Network not only achieves better accuracy than the existing methods and the more such as Deep Metric Learning but also performs significantly more efficient computation. With an accuracy of 90.2 % at a precision of 88.5%, the framework's inference time is much faster than that of Deep Metric Learning at 2.0ms. Furthermore, the proposed method performs well on the generalization over datasets with a 15% improvement of cross-domain matching accuracy on Euclidean Distance. By striking this balance between accuracy, efficiency and adaptability, it shows its suitability for dynamic and noisy environments for real-time tracking. The proposed method fills existing gaps and achieves a new state of the art for feature matching and tracking performance. © 2025 IEEE.

关键词： Object tracking

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：