检索结果-内蒙古大学图书馆

Classification of benign and malignant breast lesions in mammograms using dense-unified multiscale attention network and data-efficient image transformers

引用

The European Physical Journal Special Topics 2025年 1-19页

作者： Paavankumar, S. Karthik, R. Idayachandiran, Gangatharan Sri, Pulaparthi Penchala Deepthi Illakiya, T. School of Computer Science Engineering Vellore Institute of Technology Chennai India Centre for Cyber Physical Systems Vellore Institute of Technology Chennai India Department of Computational Intelligence Faculty of Engineering and Technology School of Computing SRM Institute of Science and Technology Chennai India

Breast cancer is a common and a serious health problem and it is the major cause of morbidity and mortality for women. Early detection of the disease is particularly challenging because abnormalities such as masses and microcalcifications exhibit subtle and diverse characteristics that are often difficult to identify in mammograms. In recent years, advancement in artificial intelligence, particularly deep learning (DL), has shown to improve diagnostic accuracy and early-stage tumor detection. This study aims to improve performance of DL models by considering both masses and microcalcifications in the proposed work to classify breast cancer abnormalities. The proposed work introduces a novel dual-track network that employs a combination of dense-unified multiscale attention fusion (UMAF) track and data-efficient image transformer (DeiT). The DeiT track processes the entire image simultaneously using patch embeddings, enabling them to capture multiscale representations and dependencies across the entire image. Simultaneously, the Dense-UMAF track focuses on extracting localized features while utilizing connectivity of DenseNet architecture to enable effective feature reuse. This approach generates relevant input features through residual connections of varying lengths, thereby effectively addressing the vanishing gradient problem. The UMAF improves feature extraction by capturing multiscale information, resulting in a better representation of the input data. This dual-track architecture is specifically designed to capture the characteristics of mass and calcification abnormalities in mammograms, which display both localized features and global contextual patterns. The proposed network was evaluated on the Curated Breast Imaging Subset of Digital Database for Screening Mammography dataset, obtaining a classification accuracy of 88.69%.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Modified Huygens Source for Maximum Gain of Implanted Antenna

Modified Huygens Source for Maximum Gain of Implanted Antenn...

引用

European Conference on Antennas and Propagation, EuCAP

作者： Jiajun Li Zvonimir Šipuš Anja K. Skrivervik Microwaves and Antennas Group École Polytechnique Fédérale de Lausanne Lausanne Switzerland Faculty of Electrical Engineering and Computing University of Zagreb Zagreb Croatia

ISBN: (数字)9788831299107

ISBN: (纸本)9798350366327

Implanted antennas are a key part of wireless implanted medical devices. However, the lossy nature of the surrounding body tissue significantly reduces their radiation efficiency and gain. Therefore, it is important to maximize the power reaching free space from the implanted antenna. The fundamental modes TE10 and TM10 are of great interest as they have lower near-field losses compared to higher order modes. A Huygens source antenna that combines two first modes, TE10 and TM10, is thus a good candidate for an efficient implanted antenna. However, because the losses for TE and TM modes differ in the implanted scenario, the optimal power distribution between the two modes will vary from that in free space.. Hence, in this contribution we study the theoretical maximum gain and the corresponding optimal power distribution of a Huygens source antenna in a deeply implanted scenario. The encapsulation size and implantation depth, the key parameters affecting the near-field efficiency, are investigated in detail. The Ohmic losses of the electric and magnetic dipole are also discussed for practical design considerations.

关键词： Wireless communication Encapsulation Medical devices Dipole antennas Antenna theory Power distribution Europe Propagation losses Bioelectric phenomena

来源：评论

学校读者我要写书评

暂无评论

Harnessing the Potential of Omnidirectional UAVs in RIS-Enabled Wireless Networks

Harnessing the Potential of Omnidirectional UAVs in RIS-Enab...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Abdoul Karim A. H. Saliah Hajar El Hammouti Daniel Bonilla Licea College of Computing Mohammed VI Polytechnic University (UM6P) Morocco Faculty of Electrical Engineering Czech Technical University Prague Czech Republic

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Multirotor Aerial Vehicles (MRAVs) when integrated into wireless communication systems and equipped with a Reflective Intelligent Surface (RIS) enhance coverage and enable connectivity in obstructed areas. However, due to limited degrees of freedom (DoF), traditional under-actuated MRAVs with RIS are unable to control independently both the RIS orientation and their location, which significantly limits network performance. A new design, omnidirectional MRAV (o-MRAV), is introduced to address this issue. In this paper, an o-MRAV is deployed to assist a terrestrial base station in providing connectivity to obstructed users. Our objective is to maximize the minimum data rate among users by optimizing the o-MRAV’s orientation, location, and RIS phase shift. To solve this challenging problem, we first smooth the objective function and then apply the Parallel Successive Convex Approximation (PSCA) technique to find efficient solutions. Our simulation results show significant improvements of 28% and 14% in terms of minimum and average data rates, respectively, for the o-MRAVs compared to traditional u-MRAVs.

关键词： Wireless networks Simulation Signal processing Terrestrial communications Planning Relays Speech processing Surface treatment Robots Optimization

来源：评论

学校读者我要写书评

暂无评论

Cervical Cancer Risk Prediction using SMOTE and Explainable AI

Cervical Cancer Risk Prediction using SMOTE and Explainable ...

引用

Pervasive Computational Technologies (ICPCT), International Conference on

作者： Priyansh Agarwal Annanya Chaudhary T. R. Sarvanan Department of Computational Intelligence School of Computing SRM Institute of Science and Technology Department of Computational Intelligence Faculty of Engineering & Tech. SRM Institute of Science and Technology

ISBN: (数字)9798331508685

ISBN: (纸本)9798331519476

Cervical cancer, accounting for 7.9% of cancers in women globally, is a critical health challenge, particularly due to its asymptomatic nature in early stages. This study proposes a machine learning-driven framework to predict cervical cancer risk using clinical data, addressing class imbalance through SMOTE and optimizing model performance via GridSearchCV. By evaluating five classifiers—K-Nearest Neighbors, Random Forest, Support Vector Classifier, Logistic Regression, and Decision Trees—this research identifies the Decision Tree model as the most robust, achieving superior precision, recall, and accuracy metrics. Additionally, Explainable AI techniques such as SHAP and LIME were employed to enhance interpretability, offering clinicians actionable insights into key risk factors driving predictions. The results demonstrate the effectiveness of machine learning in facilitating early detection and personalized treatment strategies for cervical cancer, thereby advancing its integration into clinical applications.

关键词： Measurement Accuracy Explainable AI Pipelines Predictive models Data models Decision trees Cervical cancer Random forests Overfitting

来源：评论

学校读者我要写书评

暂无评论

Harnessing Social Robotics and the Internet of Things to Reduce the Risk of Older Adults Developing Hypothermia and Dehydration 25th

Harnessing Social Robotics and the Internet of Things to Red...

引用

25th International Conference on Internet computing and IoT, ICOMP 2024, and 22nd International Conference on Embedded Systems, Cyber-physical Systems, and Applications, ESCS 2024, held as part of the World Congress in Computer Science, Computer engineering and Applied computing, CSCE 2024

作者： Sindi, Hani McCrindle, Rachel Biomedical Engineering University of Reading Reading United Kingdom Information Technology Department Faculty of Computing and Information Technology King Abdulaziz University Jeddah Saudi Arabia

ISBN: (纸本)9783031859229

Many older adults are at risk of developing hypothermia or becoming dehydrated, both of which can be dangerous conditions requiring hospitalization. This paper describes the development and evaluation of a novel support system, Hypothermia and Dehydration Advising Companion (HyDeAdCo), that integrates a social robot with off-the-shelf technologies within an IoT infrastructure, to help reduce the risk of older adults becoming hypothermic or dehydrated. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Intelligent robots

来源：评论

学校读者我要写书评

暂无评论

The Role of AI in Predictive UX Shaping the Future of Metaverse Interactions

The Role of AI in Predictive UX Shaping the Future of Metave...

引用

Computational Intelligence Approaches and Applications (ICCIAA), International Conference on

作者： Nedal Nwasra Khalil Omar Zahid Hussain Qaisar Software Engineering Department University of Petra Amman Jordan Faculty of Computing & Emerging Technologies Multan Pakistan Computer Science Department University of Petra Amman Jordan

ISBN: (数字)9798331523657

ISBN: (纸本)9798331523664

The Metaverse is defined as a collectively shared, persistent, and contextualized virtual space that will drive next-generation digital interaction facilitated by enhanced extended reality (XR) and edge computing. One of the major challenges within this development area is the proper anticipation and fulfillment of user context needs, preferences, and intentions in real-time. This paper proposes the concept of Predictive User Experience through AI for a turn to solve this challenge. With enabled predictive analytics, machine learning, and real-time data processing, the proposed AI-based predictive UX model dynamically adjusts metaverse interface features, content delivery, and modes of interaction to enhance immersion, all while maximizing user satisfaction. In this paper, we discuss AI's potential for explanatory mechanics in predictive UX and how it shapes the future of metaverse interactions. We propose a conceptual framework that describes how data collection, data processing, and adaptive interface design can be integrated within a feedback-driven architecture that seamlessly predicts and adjusts to user requirements. In this paper, we present a research methodology that integrates qualitative and technical approaches-from data modeling and architecture design up to scenario-based assessments-to validate our UX framework's predictiveness. Hence, we proceed with a use case that exemplifies the applicability of AI-based predictive UX systems for ensuring user presence and interaction consistency within a metaverse educational scenario. The discussion synthesizes and finally concludes the most pertinent findings relevant to user engagement and personalization regarding privacy and ethical considerations without compromising how these technologies can be safely and effectively scaled. We conclude our paper with the opinion that XR interactions driven by AI predictive UX mark a milestone in how applications are created, not just for entertainment and gaming but also

关键词： Adaptation models Metaverse Extended reality Machine learning Computer architecture Predictive models Data collection Data models User experience Real-time systems

来源：评论

学校读者我要写书评

暂无评论

SafeSlice: Enabling SLA-Compliant O-RAN Slicing via Safe Deep Reinforcement Learning

arXiv

引用

arXiv 2025年

作者： Nagib, Ahmad M. Abou-Zeid, Hatem Hassanein, Hossam S. School of Computing Queen’s University Canada Department of Electrical and Software Engineering University of Calgary Canada Faculty of Computers and Artificial Intelligence Cairo University Egypt

Deep reinforcement learning (DRL)-based slicing policies have shown significant success in simulated environments but face challenges in physical systems such as open radio access networks (O-RANs) due to simulation-to-reality gaps. These policies often lack safety guarantees to ensure compliance with service level agreements (SLAs), such as the strict latency requirements of immersive applications. As a result, a deployed DRL slicing agent may make resource allocation (RA) decisions that degrade system performance, particularly in previously unseen scenarios. Real-world immersive applications require maintaining SLA constraints throughout deployment to prevent risky DRL exploration. In this paper, we propose SafeSlice to address both the cumulative (trajectory-wise) and instantaneous (state-wise) latency constraints of O-RAN slices. We incorporate the cumulative constraints by designing a sigmoid-based risk-sensitive reward function that reflects the slices’ latency requirements. Moreover, we build a supervised learning cost model as part of a safety layer that projects the slicing agent’s RA actions to the nearest safe actions, fulfilling instantaneous constraints. We conduct an exhaustive experiment that supports multiple services, including real virtual reality (VR) gaming traffic, to investigate the performance of SafeSlice under extreme and changing deployment conditions. SafeSlice achieves reductions of up to 83.23% in average cumulative latency, 93.24% in instantaneous latency violations, and 22.13% in resource consumption compared to the baselines. The results also indicate SafeSlice’s robustness to changing the threshold configurations of latency constraints, a vital deployment scenario that will be realized by the O-RAN paradigm to empower mobile network operators (MNOs). Copyright © 2025, The Authors. All rights reserved.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Air Pollution Source Tracing by VIME-XGBoost

Air Pollution Source Tracing by VIME-XGBoost

引用

International Symposium on Computational and Business Intelligence (ISCBI)

作者： Wenchuan Zhao Qi Zhang Ting Shu Faculty of Data Science City University of Macau Macau China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China

ISBN: (数字)9798331533786

ISBN: (纸本)9798331533793

Air pollution source tracing, which mainly identifies and analyzes the specific sources that lead to air pollution, is a prerequisite for preventing and controlling air pollution. Traditional source analysis methods are sensitive to insufficient or inaccurate data and rely on extensive data collection, which consumes much time, labor and financial resources. Therefore, this study proposes a VIME-XGBoost two-stage model, which can learn efficient data representations from limited data and significantly enhance cross-city pollution source resolution. The model combines the self-supervised learning framework in VIME (Value Imputation and Mask Estimation) with XGBoost Regressor, and the rich features are obtained by Feature Vector Estimator and Mask Vector Estimator in VIME, which are transferred to XGBoost Regressor for nonlinear learning. Real air pollution data from five cities in China are used in the study. The average R 2 reaches 0.84, and the average prediction error of the pollution sources among the five cities is less than 1%. Comparing the experimental results with other models, VIME-XGBoost shows stability and accuracy across cities and different pollution sources and has excellent generalization ability, providing a powerful and effective solution for pollution source analysis and environmental data prediction.

关键词： Atmospheric modeling Urban areas Self-supervised learning Machine learning Predictive models Air pollution Vectors Data models Stability analysis Imputation

来源：评论

学校读者我要写书评

暂无评论

Faceswap Finder: A Fusion-Based Deepfake Detection Technique

Faceswap Finder: A Fusion-Based Deepfake Detection Technique

引用

International Conference on Advancements in Computational Sciences (ICACS)

作者： Nimra Basit Fatima Khalid Qurat Ul Ain Maria Andleeb Department of Software Engineering University of Engineering and Technology Taxila Pakistan Faculty of Computer Sciences and Engineeing GIKI Topi Pakistan Department of Computing Shifa Tameer-e-Millat University PRC Islamabad Pakistan

ISBN: (数字)9798331533038

ISBN: (纸本)9798331533045

The significant improvement in deepfake generation algorithms has made possible the manipulation of visual data. The advancement of GAN technology and tools has made it relatively simple to use source and target photos and build realistic deepfakes. The challenges associated with deepfakes include political defamation, impersonation, misinformation, and cyberstalking. The generation of deepfakes through faceswap approach hinders the detection oftentimes as it spans numerous artifacts – illumination, facial skin tones, and other conditions. Existing techniques do not employ robust mechanisms to detect such deepfakes either due to computational inefficiency or focusing on a few parameters for decision-making. To counteract these dangers, a robust deepfake detection model must be developed and implemented. This paper proposes Efficient-Fused Net (EF-Net) - a transfer learning approach to two EfficientNet-B5 models with partial freezing between its layers that can differentiate between real and fake images generated from faceswap implementation. We evaluated our model’s performance using the challenging deepfake subset of the extensive and varied FaceForensics++ and DFDC-Preview datasets. The proposed method outperforms state of the art methods in the detection of faceswap deepfakes, surpassing the performance by $89.19 \%$ accuracy on DFDC-Preview and $\mathbf{9 7 . 7 0 \%}$ of Deepfakes subset of FaceForensics++.

关键词： Deepfakes Visualization Computational modeling Transfer learning Decision making Lighting Focusing Skin Robustness Resilience

来源：评论

学校读者我要写书评

暂无评论

Mitigating Poisoning Attacks in Federated Learning Through Deep One-Class Classification

引用

IEEE Transactions on Cognitive Communications and Networking 2025年

作者： Zhang, Anqi Zhao, Ping Lu, Wenke Zhou, Yipeng Zhang, Wenqian Zhang, Guanglin Donghua University College of Information Science and Technology Shanghai201620 China Ministry of Education Engineering Research Center of Digitized Textile and Apparel Technology Shanghai201620 China Macquarie University School of Computing Faculty of Science and Engineering 2122 Australia

Federated learning (FL) enables clients to learn a machine learning model collaboratively without sharing their private local data to the server. However, due to its distributed structure, FL is vulnerable to poisoning attacks where adversaries intentionally send the poisoned local model parameters to the server and further affect the behavior of the global model. Existing works on mitigating poisoning attacks in FL are difficult to accurately characterize data relationships of high-dimensional parameters. Moreover, they cannot guarantee Byzantine robustness when the majority of clients are compromised (Byzantine ratio) and clients’ local datasets are highly non-independent and identically distributed (non-IID). In this paper, we conduct a pioneering work to introduce deep one-class classification into mitigating poisoning attacks in FL, which can guarantee Byzantine robustness even under a Byzantine ratio and non-IID degree greater than 0.5 with the assistance of a root dataset. Our key idea is to adequately learn vector features of benign local model parameters using Deep Support Vector Data Description (Deep SVDD) and achieve the optimal classification by training a deep learning-based one-class classifier equipped with a proper decision boundary based on the root dataset. To further optimize the classifier, we employ a regularizer based on random noise injections, which can address the hypersphere collapse problem inherent in Deep SVDD. Exhaustive experiments on MNIST, F-MNIST, and CIFAR-10 demonstrate that compared with five typical Byzantine-robust methods, our defense strategy achieves excellent effectiveness in mitigating targeted/untargeted poisoning attacks and an adaptive attack in FL. Even with a similar but different root dataset, it still maintains a good Byzantine robustness. © 2015 IEEE.

关键词： Federated learning

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：