检索结果-内蒙古大学图书馆

5th International Conference on Advanced Research in Computing, ICARC 2025

作者： Yeshani, R.K.B. Chiththranjan, Charith University of Moratuwa Dept. of Computer Science and Eng. Sri Lanka

ISBN: (纸本)9798331530983

This work presents an innovative approach to image-to-image translation, focusing on converting colour pencil images, grayscale pencil images, and grayscale images into RGB images using the Pix2Pix GAN. Building upon the established Pix2Pix framework, which utilizes a U-Net based architecture with skip connections for the generator and a Patch-GAN discriminator. This aims to refine the translation process, addressing the images' intricate details and colour dynamics. Empirical results demonstrate that this generator architecture significantly outper- forms the U-Net architecture in producing more vivid, detailed and accurate RGB images. By leveraging this method, the study not only contributes to the body of knowledge in neural network applications for artistic image translation but also opens new avenues for further researches in image translations. This work underscores the potential of targeted architectural modifications within GAN frameworks to achieve image translation outcomes, thereby setting a precedent for future explorations in the field. © 2025 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Classifying white blood cells using combining different convolutional neural networks

引用

Multimedia Tools and Applications 2025年 1-24页

作者： Toptaş, Murat Toptaş, Buket Hanbay, Davut Software Eng. Dept. Engineering and Natural Science Faculty Bandırma Onyedi Eylül University Balıkesir Turkey Computer Eng. Dept. Engineering Faculty Inonu University Malatya44280 Turkey

White blood cells are warrior cells that protect the human body against external factors. Each of these warrior cells performs a distinct task, making every piece of information about them highly valuable in the medical field. In this article, a classification framework for the four known types of white blood cells is proposed. It is hoped that the classification of these types will contribute to the prediction of diseases such as AIDS, malaria, leukemia, and many others. In the proposed method, images of white blood cells from the Blood Cell Classification and Detection dataset were used as input to Convolutional Neural Networks. The feature vectors extracted using these Convolutional Neural Network architectures were combined into a single vector. A Minimum Redundancy Maximum Relevance algorithm was then employed to identify the most effective features within the feature vector. Experiments were conducted using these selected features, and the analysis of each experiment was reported in detail. The Support Vector Machines classifier achieved an accuracy of 98.63% in classifying white blood cell types by combining features from multiple deep learning architectures. The experimental results demonstrated that the features obtained from different layers of the Convolutional Neural Networks had varying impacts on the classification performance. © The Author(s) 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

SmartPattern: A Machine Learning Framework for Detecting Reentrancy Vulnerabilities in Blockchain Smart Contracts 59

SmartPattern: A Machine Learning Framework for Detecting Ree...

引用

59th Annual Conference on Information sciences and Systems, CISS 2025

作者： Ajibola, Temitope Ezekiel Adeniran, Opeyemi Taiwo Taiwo, Peter Morgan State University Dept. of Electrical and Computer Eng. Baltimore United States

ISBN: (纸本)9798331513269

This study introduces SmartPattern, a novel machine learning-based framework to detect reentrancy attacks in smart contracts, a critical threat to blockchain security. Analyzing 40,000 smart contract, SmartPattern achieves 94% detection accuracy with Random Forest and Support Vector Classifier, outperforming Bidirectional Encoder Representations Transformers embeddings, which produced inconsistent accuracies of 84% and 78% Random Forest and Support Vector Classifier, respectively. Unlike traditional tools like Slither, which rely on static analysis and predefined heuristics, SmartPattern overcomes limitations related to dynamic invocation patterns and non-linear state changes. By leveraging machine learning models and targeted pattern recognition, SmartPattern effectively detects obfuscated attack patterns and generalizes to unseen vulnerabilities. This scalable, automated framework significantly enhances blockchain security by safeguarding billions of dollars in digital assets and promoting trust in decentralized ecosystems. The results demonstrate that SmartPattern is a viable alternative to state-of-the-art models, including those using Graph Convolutional Networks, and provides a comprehensive solution for fortifying smart contract ecosystems against reentrancy attacks. © 2025 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Estimation of Multi-Attribute Differential Graphs with Non-Convex Penalties

Estimation of Multi-Attribute Differential Graphs with Non-C...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Tugnait, Jitendra K. Dept. of Electrical & Computer Eng. Auburn University AuburnAL36849 United States

ISBN: (纸本)9798350368741

We consider the problem of estimating differences in two multi-attribute Gaussian graphical models (GGMs) which are known to have similar structure, using a penalized D-trace loss function with nonconvex penalties. The GGM structure is encoded in its precision (inverse covariance) matrix. Existing methods for multi-attribute differential graph estimation are based on a group lasso penalized loss function. In this paper, we consider a penalized D-trace loss function with nonconvex (log-sum and smoothly clipped absolute deviation (SCAD)) penalties. Two proximal gradient descent methods are presented to optimize the objective function. Theoretical analysis establishing local consistency in support recovery, local convexity and estimation in high-dimensional settings is provided. We illustrate our approach with a numerical example. © 2025 IEEE.

关键词： ifferential graph learning log-sum multiattribute graphs SCAD penalties undirected graph

来源：评论

学校读者我要写书评

暂无评论

Optimal Protection Coordination for Mitigating Sympathetic Tripping in DER-Integrated Power Systems

Optimal Protection Coordination for Mitigating Sympathetic T...

引用

2025 IEEE Texas Power and Energy Conference, TPEC 2025

作者： Almoola, Ahmed Ahmed, Sara Gatsis, Nikolaos The University of Texas at San Antonio Dept. of Electrical & Computer Eng. San AntonioTX United States

ISBN: (纸本)9798331541125

The integration of distributed energy resources (DERs) is revolutionizing the power grid but introduces nu-merous challeng.s for traditional protection schemes. These include blinding, selectivity issues, sensitivity reductions, and miscoordination. Among these, sympathetic tripping, whereby relays respond to faults in adjacent zones despite no fault in their own zone, has emerged as a critical issue, compromising grid reliability and causing avoidable outages. This paper focuses on addressing sympathetic tripping in DER-integrated networks. A novel mathematical optimization model is developed with pertinent binary and slack variables to respectively model sympathetic trip conditions and differences between relay operation times. The optimization model produces relay settings that coordinate primary-backup relay pairs and mitigate sympathetic tripping instances. Validated on the IEEE 123-bus distribution system, the approach demonstrates reduction in sympathetic tripping, improving protection coordination in DER-rich environments. © 2025 IEEE.

关键词： Overcurrent relay relay misoperation

来源：评论

学校读者我要写书评

暂无评论

Benchmarking Edge AI Platforms: Performance Analysis of NVIDIA Jetson and Raspberry Pi 5 with Coral TPU

Benchmarking Edge AI Platforms: Performance Analysis of NVID...

引用

2025 IEEE SoutheastCon, SoutheastCon 2025

作者： Minott, David Siddiqui, Salman Haddad, Rami J. Georgia Southern University Dept. of Electrical and Computer Eng. StatesboroGA United States

ISBN: (纸本)9798331504847

The rapid development of edge AI applications has led to the demand for high-performance, power-efficient embedded computing platforms. This paper evaluates the computational efficiency, power consumption, and thermal characteristics of the NVIDIA Jetson Orin NX, Jetson Nano, and Raspberry Pi 5 with Coral TPU across various AI workloads. Performance metrics, including inference speed, CPU and GPU utilization, and thermal stability, were benchmarked using a YOLOv5 object detection model. The results indicate that the Jetson Orin NX achieves the highest frame rates, nearly doubling the performance of the Raspberry Pi 5 with Coral TPU (41.8 FPS vs. 21.5 FPS) and surpassing the Jetson Nano by more than 50 percent. Power efficiency was another critical factor, with the Jetson Nano consuming the least power at approximately 7W, while the Raspberry Pi 5 with Coral TPU consumed 8.3W and the Jetson Orin NX reached 10.6W under load. Thermal analysis revealed that the Raspberry Pi 5 operates at significantly higher temperatures, reaching up to 80°C without active cooling, whereas the Jetson Nano and Jetson Orin NX maintained stable temperatures at 42°C and 45°C, respectively. These findings underscore the importance of active cooling solutions and optimized workload distribution for edge AI devices. The comparative assessment provided in this study helps guide hardware selection for specific edge AI applications, balancing trade-offs between power efficiency, thermal performance, and computational speed. © 2025 IEEE.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

Federated Semi-Supervised Medical Image Classification Using Improved Inter-Client Relation Matching 8

Federated Semi-Supervised Medical Image Classification Using...

引用

8th International Conference on Data science and Machine Learning Applications, CDMA 2025

作者： Ozdemir, Muhammed Said Sabbagh, Mais Erdem, Cigdem Eroglu Korcak, Omer Uludag, Kubra Marmara University Faculty of Eng. Department of Computer Engineering İstanbul Turkey Ozyegin University İstanbul Turkey Dept. of Electrical and Electronics Eng. Istanbul Turkey Marmara University Faculty of Engineering Department of Computer Engineering İstanbul Turkey

ISBN: (纸本)9798331539696

Traditional centralized machine learning faces chal-leng.s with 'data label scarcity' and 'data privacy', especially in medical data classification. Semi-supervised learning leverages unlabeled data, while federated learning protects data privacy by decentralizing model training. In this study, we introduce an improved inter-client relation matching algorithm for semi-supervised federated medical image classification (iFedIRM). The FedIRM algorithm defines a loss function to transfer the knowledge of disease relationships from labeled clients to unlabeled clients to extract discriminative information from unlabeled data using per-category mean feature vectors. Our approach enhances the FedIRM algorithm by incorporating per-category feature covariance matrices for better feature representation and using group-based averaging for model aggregation. We utilize the ConvNeXt architecture as the backbone, and apply a confidence threshold to filter unreliable pseudo-labels. Experiments on three medical datasets show improved accuracy, F1, and AUC metrics. Code will be available at https://***/msoz7/iFedIRM. © 2025 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Non-negative Weighted DAG Structure Learning

Non-negative Weighted DAG Structure Learning

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Rey, Samuel Saboksayr, Seyed Saman Mateos, Gonzalo Dept. of Signal Theory and Communications Rey Juan Carlos University Madrid Spain Dept. of Electrical and Computer Eng. University of Rochester RochesterNY United States

ISBN: (纸本)9798350368741

We address the problem of learning the topology of directed acyclic graphs (DAGs) from nodal observations, which adhere to a linear structural equation model. Recent advances framed the combinatorial DAG structure learning task as a continuous optimization problem, yet existing methods must contend with the complexities of non-convex optimization. To overcome this limitation, we assume that the latent DAG contains only non-negative edge weights. Leveraging this additional structure, we argue that cycles can be effectively characterized (and prevented) using a convex acyclicity function based on the log-determinant of the adjacency matrix. This convexity allows us to relax the task of learning the non-negative weighted DAG as an abstract convex optimization problem. We propose a DAG recovery algorithm based on the method of multipliers, that is guaranteed to return a global minimizer. Furthermore, we prove that in the infinite sample size regime, the convexity of our approach ensures the recovery of the true DAG structure. We empirically validate the performance of our algorithm in several reproducible synthetic-data test cases, showing that it outperforms state-of-the-art alternatives. © 2025 IEEE.

关键词： causal discovery convex relaxation DAG learning graph signal processing network topology inference

来源：评论

学校读者我要写书评

暂无评论

Towards speed-of-sound imaging with conventional ultrasound transducers using laser diode photoacoustic

Towards speed-of-sound imaging with conventional ultrasound ...

引用

Medical Imaging 2025: Ultrasonic Imaging and Tomography

作者： Bezek, Can Deniz Moradi, Hamid Rohling, Robert Salcudean, Septimiu Goksel, Orcun Department of Information Technology Uppsala University Uppsala Sweden Dept. of Electrical & Computer Eng. University of British Columbia VancouverBC Canada

ISBN: (纸本)9781510686021

Most imaging solutions that utilize ultrasound wave propagation require the projection of temporal signals received by transducer elements into spatial maps. This process, known as beamforming, requires knowledge of the speed-of-sound (SoS) in the medium. Incorrect SoS assumptions lead to aberration artifacts, reducing image quality and limiting clinical usability. SoS is also a novel imaging biomarker for assessing tissue characteristics. In this study, we propose a spatial SoS distribution estimation method using conventional hand-held ultrasound transducers and simple laser-diode based photoacoustics. By identifying the time of flight between photoacoustic point sources and transducer elements and relating these to SoS values along corresponding propagation paths, the underlying SoS maps are reconstructed by solving an inverse problem. We validate our method through numerical simulations and ex vivo experiments. Numerical phantoms are successfully reconstructed under various noise levels and numbers of photoacoustic sources. In ex vivo experiments with chicken breast, the estimated SoS value is consistent with results reported in the literature. The proposed approach offers a low-cost, compact solution for photoacoustics-based SoS estimation in various clinical applications, such as breast and intra-operative prostate imaging, both for diagnosis and for improving image quality of acoustic-based modalities, including photoacoustic imaging. © 2025 SPIE.

关键词： Ultrasonic propagation

来源：评论

学校读者我要写书评

暂无评论

Color and Grayscale Pencil Image Translation Into Realistic Images Using Generative Adversarial Networks

Color and Grayscale Pencil Image Translation Into Realistic ...

引用

Advanced Research in Computing (ICARC), International Conference on

作者： R. K. B. Yeshani Charith Chiththranjan Dept. of Computer Science and Eng. University of Moratuwa Sri Lanka

ISBN: (数字)9798331530983

ISBN: (纸本)9798331530990

This work presents an innovative approach to image-to-image translation, focusing on converting colour pencil images, grayscale pencil images, and grayscale images into RGB images using the Pix2Pix GAN. Building upon the established Pix2Pix framework, which utilizes a U-Net based architecture with skip connections for the generator and a Patch-GAN discriminator. This aims to refine the translation process, addressing the images’ intricate details and colour dynamics. Empirical results demonstrate that this generator architecture significantly outper- forms the U-Net architecture in producing more vivid, detailed and accurate RGB images. By leveraging this method, the study not only contributes to the body of knowledge in neural network applications for artistic image translation but also opens new avenues for further researches in image translations. This work underscores the potential of targeted architectural modifications within GAN frameworks to achieve image translation outcomes, thereby setting a precedent for future explorations in the field.

关键词： Knowledge eng.neering Translation Image color analysis Architecture Neural networks Focusing computer architecture Gray-scale Generative adversarial networks Generators

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：