检索结果-内蒙古大学图书馆

Optimized Convolutional Neural Network at the IoT edge for image detection using pruning and quantization

Multimedia Tools and Applications 2025年第9期84卷 5435-5455页

作者： Naveen, Soumyalatha Kounte, Manjunath R Department of Computer Science and Engineering Manipal Institute of Technology Bengaluru Manipal Academy of Higher Education Manipal India School of Electronics and Communication Engineering REVA University Bangalore India

Most real-time computer vision applications heavily rely on Convolutional Neural Network (CNN) based models, for image classification and recognition. Due to the computationally and memory-intensive nature of the CNN model, it’s challenging to deploy on resource-constrained Internet of Things (IoT) devices to enable Edge intelligence for real-time decision-making. Edge intelligence requires minimum inference latency, memory footprint, and energy-efficient model. This work aims to develop an energy-efficient deep learning accelerator using a 3-stage pipeline: Training, Weight-pruning, and Quantization to reduce the model size and optimize the resources. First, we employ YOLOv3, a CNN architecture to detect objects in an image on the trained data. In addition, a sparse network of YOLO has been created by using pruning, which helps to improve the network’s performance and efficiency by reducing the computational requirements. Finally, we utilize 8-bit quantization to reduce the precision of the weights and activations, in a neural network. The evaluation of our proposed model shows that combining pruning and 8-bit quantization improves the efficiency and performance of the model. While pruning shows a decline of 80.39% in model parameters. The combination of 8-bit quantization results in an improvement in inference latency by 22.72% compared to existing SQuantization approach and a reduction of energy consumption by 29.41%. © The Author(s) 2024.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Deep ensemble optimal classifier-based software defect prediction for early detection and quality assurance

Journal of Umm Al-Qura University for Engineering and Archit...

引用

Journal of Umm Al-Qura University for engineering and Architecture 2025年 1-12页

作者： Srinivas, Kalyanapu Janapati, Krishna Chaithanya Sai Kiran, Mungalched Sudheer Reddy, Davu Vinay, Thota Vamshi Yashwanth, Choula Department of Computer Science and Engineering Dhanekula Institute of Engineering and Technology Andhra Pradesh Vijayawada India Department of Electronics and Communication Engineering Vardhaman College of Engineering Telangana Hyderabad India Department of Computer Science and Engineering (Data Science) Malla Reddy Engineering College and Management Sciences Medchal Telangana Kistapur India

Software Defect Prediction (SDP) is critical in identifying fault-prone modules during the software development life cycle, enhancing software quality, and reducing maintenance costs. However, existing SDP models face significant challenges, including high false positive rates and severe class imbalance between defect-prone and non-defect modules, which hinder practical model training and accurate defect detection. To overcome these challenges, this study proposes a robust and efficient SDP architecture that integrates several advanced techniques to enhance predictive accuracy and model reliability. The methodology begins with comprehensive data preprocessing, incorporating normalization to standardize feature scales and outlier detection to eliminate noise and improve data quality. To address the inherent class imbalance in the dataset, the minority oversampling by synthetic data (MOSD) technique is employed to generate synthetic samples for the minority class, thus achieving a more balanced distribution. Following this, an Adaptive sequential K-best (ASKB) feature selection strategy is implemented to identify and retain the most relevant features, effectively reducing dimensionality while preserving critical information. Finally, the weighted random forest (WRF) classifier is utilized, which assigns appropriate weights to training instances, enabling improved classification performance, especially for minority class instances. The proposed methodology achieved an accuracy of 99.119%, a precision of 99.431%, a recall of 99.122%, and an F1-score of 99.333%. © The Author(s) 2025.

关键词： Adaptive sequential K-best Class imbalance Data preprocessing Feature selection Minority oversampling Predictive modeling Software defect prediction Synthetic data Weighted random forest

来源：评论

学校读者我要写书评

暂无评论

Autoencoder-Based Architecture for Identification and Mitigating Phishing URL Attack in IoT Using DNN

引用

Journal of The Institution of Engineers (India): Series B 2023年第6期104卷 1227-1240页

作者： Gopal, S.B. Poongodi, C. Nanthiya, D. Kirubakaran, T. Kulavishnusaravanan, B. Logeshwar, D. Department of Electronics and Communication Engineering Kongu Engineering College Perundurai India Department of Computer Science and Engineering Vivekanandha College of Engineering for Women Namakkal India Department of Computer Technology-UG Kongu Engineering College Perundurai India

As a new and potentially devastating form of cyberattack, ‘Phishing’ URLs pose a risk to users by impersonating legitimate websites in an effort to obtain sensitive information such as usernames, passwords, and financial details. At least two-thirds of persons were hit by these phishing assaults in 2016, according to a survey (Gopal et al. in IOP Conference Series: Materials science and engineering, vol. 1055, no. 1, p. 012072. IOP Publishing). Therefore, these attacks must be addressed carefully to prevent financial losses. In order to detect and prevent new kinds of assault, in addition to the ones already in use, a defense mechanism that can learn the possibilities of attacks is required. In the IoT (Internet of Things), everything is managed remotely via a web address (URL). So, intruders can simply gain access to the controls by stealing the URL used to access them from a distant location. To identify and prevent access to phishing sites at the network layer, we offer an approach that combines a deep neural networks (DNN) model with autoencoders. In order to train the model for feature reduction, an autoencoder sends it a dataset that includes both legitimate and phishing domains to analyze. The effectiveness of the suggested approach has been tested on a number of datasets, including Open phish, UCI, Mendeley, and PhishTank. For the multiclassification dataset, these metrics for this design are: 92.89% accuracy, 93.07% recall, 92.75% precision, and 92.21% F1-score. © 2023, The Institution of Engineers (India).

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

EEG Emotion Prediction Using Artificial Intelligence

EEG Emotion Prediction Using Artificial Intelligence

引用

2024 International Conference on Signal Processing, Computation, electronics, Power and Telecommunication, IConSCEPT 2024

作者： Singh, Samvardhan Chaudhary, Kshitij Devi, Durga SRM institute of science and technology Electronics and computer engineering Chennai India SRM institute of science and technology Computer science engineering Chennai India

ISBN: (纸本)9798331540685

Emotions are fundamental for human beings and play an important role in human cognition. Emotion is commonly associated with logical decision making, perception, human interaction, and to a certain extent, human intelligence itself. Emotion detection discussed in this paper is a technique that uses electroencephalography (EEG) to measure brain activity and identify different emotional states, it is a non-invasive method that measures electrical activity in the brain through electrodes placed on the scalp. The electrical activity of the brain is related to different cognitive and emotional processes, and electroencephalography can be used to measure changes in brain activity that occur in response to different emotions. The emotion prediction model can be used for early detection and of mental health disorders and to tailor personalised treatment plans for such patients, to help researchers gain a better understanding of the emotional aspects of various conditions and even in stress management, wellness program and more. © 2024 IEEE.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

"Survey and analysis of blockchain technologies with respect to: properties, algorithms, architecture, models, evolution and framework"

引用

Multimedia Tools and Applications 2025年第14期84卷 13571-13615页

作者： Tiwari, Nikita Ranjan, Prakash Biswal, Pradeep Kumar Barde, Chetan Sinha, Neha Department of Computer Science and Engineering Indian Institute of Information Technology Bhagalpur Bhagalpur India Department of Electronics and Communication Engineering Indian Institute of Information Technology Bhagalpur Bhagalpur India

The growing realm of blockchain technology has captivated researchers and practitioners alike with its promise of decentralized, secure, and transparent transactions. This paper presents a comprehensive survey and analysis of various blockchain technologies, focusing on their fundamental properties, architectural designs, operational models, evolutionary trends, and overarching frameworks. Through an extensive review of existing literature and empirical studies, this research synthesizes the multifaceted landscape of blockchain, elucidating key concepts such as consensus mechanisms, smart contracts, scalability, interoperability, and privacy protocols. By examining the evolution of blockchain from its inception to its current state and beyond, this paper delineates the trajectory of technological advancements and identifies emerging paradigms. Furthermore, it critically evaluates the strengths and limitations of diverse blockchain architectures and models, offering insights into their applicability across different domains and use cases. Finally, this study culminates in the proposal of a comprehensive framework that encapsulates the essential elements necessary for understanding, implementing, and harnessing the transformative potential of blockchain technologies. Through this holistic analysis, this research contributes to a deeper understanding of blockchain ecosystems and provides valuable guidance for future research directions and practical implementations. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Blockchain

来源：评论

学校读者我要写书评

暂无评论

Performance of quantum key distribution in free-space optical systems under various atmospheric channel effects

引用

Optical engineering 2025年第4期64卷

作者： Ata, Yalçın Kiasaleh, Kamran OSTIM Technical University Department of Electrical and Electronics Engineering Ankara Turkey University of Texas at Dallas Erik Jonsson School of Engineering and Computer Science RichardsonTX United States

We devoted to analyzing information transfer in quantum channels in terms of the performance of quantum key distribution (QKD) free-space optical (FSO) communication systems through Bennett and Brassard 84 (BB84) protocol by taking into account the combined effect of attenuation, turbulence, pointing error, and angle of arrival (AOA) fluctuations. Path loss is modeled by Beer-Lambert law, and the turbulent channel is assumed to be Fisher-Snedecor F distributed. The pointing error and AOA fluctuations are modeled by Hoyt and Rayleigh distributions, respectively. The error and sift probabilities, quantum bit error rate, and secret key rate of a QKD FSO system are obtained in analytical form, and results are presented depending on various parameters. It is shown that increased beam waist, beam deviation, AOA fluctuations, and propagation distance cause performance degradation and less secure information transmission. However, an increase in the ratio of horizontal and vertical beam misalignments, field of view, average photon number per pulse, visibility of the atmospheric environment, and optical beam wavelength impact the performance of QKD FSO systems positively. The effect of the turbulence structure constant is observed to be modest on the performance of the QKD FSO system. © 2025 Society of Photo-Optical Instrumentation Engineers (SPIE)

关键词： Space optics

来源：评论

学校读者我要写书评

暂无评论

A Compact Filtering Antenna System with Wide-Angle Scanning Capability for V2I Communication

引用

Chinese Journal of electronics 2024年第2期33卷 516-526页

作者： Chuang HAN Tong LI Zhaolin ZHANG Ling WANG Guangwei YANG School of Electronics and Information Northwestern Polytechnical University School of Electronic Engineering and Computer Science Queen Mary University of London

A compact filtering antenna system with wide-angle scanning is proposed for vehicle to infrastructure(V2I) communication which would handle complex communication scenarios. In this work, a wide beam filtering antenna is realized by using some inductive resistance structures such as metal pins and pillars, and capacitive structures such as slots, parasitical patches to produce the radiation nulls at two sides of the operating frequency band and improve the impedance matching in the passband. Meanwhile, the wide beam capability is also realized by the above structure. Furthermore, two H-and E-plane linear arrays are designed for the beam scanning capability with filtering characteristics based on the proposed antenna. To verify the proposed design concept, a prototype is fabricated and measured. The measurement and simulation agree well, demonstrating an excellent filtering characteristic with the operating frequency band from 3.18 to 3.45 GHz(about 8.1%), the high total efficiency of about 88%, and 3-d B-beamwidth of more than 100° and 120° in the above two arrays, respectively. Additionally, the proposed arrays can realize the beam scanning up to the coverage of 112° and 120° with a lower gain reduction and a good filtering characteristic, respectively.

关键词： Filtering antenna Wide beam Beam scanning Beam steering Phased array

来源：评论

学校读者我要写书评

暂无评论

Classification of suspected objects and severity assessment of COVID-19 from chest X-ray images using deep transfer learning

引用

Research on Biomedical engineering 2023年第3期39卷 705-718页

作者： Verma, Karan Kumar, Ashok Swaraj, Aman Sagar, Aditi Department of Computer Science and Engineering National Institute of Technology Delhi Delhi India Department of Electronics and Communication Engineering Government Mahila Engineering College Rajasthan Ajmer India Department of Computer Science and Engineering Indian Institute of Technology Roorkee India

Purpose: Coronavirus disease 2019 (COVID-19) has infected about 418 million people across the globe. So, the analysis of biomedical imaging accompanied with artificial intelligence (AI) approaches has transpired a vital role in diagnosing COVID-19. Until now, numerous classification approaches have been demonstrated for the detection of COVID-19. The assessment of COVID-19 patients according to severity level is not so far investigated. For this motivation, the classification of COVID-19 chest X-ray (CXR) images according to severity of the infection is presented in this work. Methods: Primarily, the 1527 CXR images are pre-processed to reshape images into unique size, denoised, and enhanced images through median filter and histogram equalization (HE) techniques, respectively. Afterward, reshaped, denoised, and enhanced CXR images are augmented using synthetic minority oversampling technique (SMOTE) to achieve the balanced dataset of 1752 CXR images. After augmentation, a pre-trained VGG16 and residual network 50 (Resnet50) deep transfer learning models with random forest (RF) and support vector machine (SVM) classifiers are utilized for feature extraction and classification of 1752 CXR images into diverse class labels such as normal, severe COVID-19, and non-severe COVID-19. Results: Our proposed ResNet50 model with SVM classifier provides the highest accuracy of about 95% for severity assessment and classification of COVID-19 CXR images as compared to other permutations. For the ResNet50 model with SVM classifier model, the average value of precision, recall, and F1-score are 91%, 94%, and 92%, respectively. Conclusion: The multi-class classification deep transfer learning models are presented to determine the severity assessment and classification of COVID-19 by using CXR images. Out of these proposed models, the ResNet50 model with SVM classifier will be highly favorable for doctors to classify patients according to their severity assessment and detection of COV

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

An efficient single unit for virtual-machine placement in cloud data centres

引用

International Journal of Information and Communication Technology 2024年第1期25卷 1-24页

作者： Ismaeel, Salam Miri, Ali Al-Khazraji, Ayman Department of Computer Science Ryerson University Toronto Canada Department of Electrical and Electronics Engineering University of Bahrain Manama Bahrain

There are numerous energy minimisation plans that are adopted in today’s data centres (DCs). The highest important ones are those that depend on switching off unused physical machines (PMs). This is usually done by optimal distribution and/or reallocating of virtual machines (VMs) on the selected servers, while maintaining the quality of service (QoS) to ensure the performance of a DC. In this work, a novel server machine condition index (MCI) has been proposed, which includes all resources related to servers available in the DC using a single unit. The MCI represents a dynamic tool to compare services, increase effectiveness, reflect PM adequation, and ensure the optimal management of heterogeneous DC resources. The MCI will be used to convert the multi-objective VM allocation optimisation problem into a single-objective problem. This work will identify the MCI components and the way that can be used as a cloud resource unit, and modified VMP algorithms. Copyright © 2024 Inderscience Enterprises Ltd.

关键词： Electric power utilization

来源：评论

学校读者我要写书评

暂无评论

Align before Attend: Aligning Visual and Textual Features for Multimodal Hateful Content Detection 18

Align before Attend: Aligning Visual and Textual Features fo...

引用

18th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2024 - Student Research Workshop, SRW 2024

作者： Hossain, Eftekhar Sharif, Omar Hoque, Mohammed Moshiul Preum, Sarah M. Department of Computer Science and Engineering Department of Computer Science Dartmouth College United States Department of Electronics and Telecommunication Engineering Chittagong University of Engineering & Technology Bangladesh

ISBN: (纸本)9798891760905

Multimodal hateful content detection is a challenging task that requires complex reasoning across visual and textual modalities. Therefore, creating a meaningful multimodal representation that effectively captures the interplay between visual and textual features through intermediate fusion is critical. Conventional fusion techniques are unable to attend to the modality-specific features effectively. Moreover, most studies exclusively concentrated on English and overlooked other low-resource languages. This paper proposes a context-aware attention framework for multimodal hateful content detection and assesses it for both English and non-English languages. The proposed approach incorporates an attention layer to meaningfully align the visual and textual features. This alignment enables selective focus on modality-specific features before fusing them. We evaluate the proposed approach on two benchmark hateful meme datasets, viz. MUTE (Bengali code-mixed) and MultiOFF (English). Evaluation results demonstrate our proposed approach’s effectiveness with F1-scores of 69.7% and 70.3% for the MUTE and MultiOFF datasets. The scores show approximately 2.5% and 3.2% performance improvement over the state-of-the-art systems on these datasets. Our implementation is available at https://***/eftekhar-hossain/Bengali-Hateful-Memes. © 2024 Association for Computational Linguistics.

关键词：

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：