检索结果-内蒙古大学图书馆

DXDSENet-CM model: an ensemble learning model based on depthwise Squeeze-and-Excitation ConvMixer architecture for the classification of multi-class skin lesions

引用

Multimedia Tools and Applications 2025年第12期84卷 9903-9938页

作者： Fırat, Hüseyin Üzen, Hüseyin Department of Computer Engineering Faculty of Engineering Dicle University Diyarbakır Turkey Department of Computer Engineering Faculty of Engineering and Architecture Bingol University Bingol Turkey

Skin cancer is acknowledged as the most prevalent form of cancer on a global scale. Failure to detect it in its initial phases can lead to fatality, underscoring the significance of early diagnosis. While visible to the naked eye, the considerable similarities between different classes and the variations within a single class create substantial hurdles for accurate detection. Given the widespread occurrence of this ailment, numerous deep learning-driven automated systems have been created to aid dermatologists in promptly identifying skin anomalies. Within this investigation, an ensemble learning architecture based on Xception, DenseNet201 and Depthwise Squeeze-and-Excitation Network and ConvMixer (DSENet-ConvMixer) architectures is presented for skin lesion classification. Xception, a pre-trained CNN network, utilizes depthwise separable convolutions and residual connections, offering efficient feature extraction from skin lesion images. DenseNet201’s parameter efficiency, feature reuse, and multi-scale information capture make it a valuable addition. DSENet-ConvMixer integrates these components, efficiently merging low-level and high-level features to enhance training and network efficiency. ConvMixer excels at feature extraction, enhancing spatial detail capture. ISIC 2018 and 2019 datasets were employed to assess the classification performance of the proposed architecture. Individual accuracies for Xception, DenseNet201, and DSENet-ConvMixer in ISIC 2018 were 85.91%, 85.71%, 82.02%, respectively, and in ISIC 2019 were 80.93%, 80.50%, 79.08%, respectively. However, the ensemble architecture with these three architectures achieved 88.21% for ISIC 2018 and 92.34% for ISIC 2019, demonstrating superior performance. Literature reviews on both datasets also confirmed the proposed architecture’s effectiveness compared to recent studies. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Implementation of Safety-Related Intrusion Detection for an Industrial Robotic Arm

Implementation of Safety-Related Intrusion Detection for an ...

引用

2025 International Conference on Control, Automation, and Instrumentation, IC2AI 2025

作者： Haydar, Georgio Abou Telawi, Samer Borcsok, Josef Architecture and System Programming University of Kassel Department of Computer Kassel Germany

ISBN: (纸本)9798331529604

Ensuring the safety of human workers collaborating with industrial robots is paramount. This research work presents a novel approach by developing a safety-related intruder detection system for the operational zones of industrial robots, following the IEC 61508 standard and adopting the related redundant architecture. The hardware integrates light curtains and radar sensors in a dual-layer configuration, while the software utilizes a safety system-on-chip to implement real-time diagnostics and fault tolerance. Protecting employees and ensuring a smooth and safe work environment are significant concerns for all companies. As a result, numerous standards and regulations have been established to provide safe collaboration between humans and machines. However, despite these measures, accidents continue to occur, resulting in varying degrees of injuries. The proposed system is depicted through a comprehensive schematic, emphasizing the integration of safety measures within the robotic environment. The detailed prototype demonstrates practical application, show-casing the system's ability to detect intrusions, initiate emergency responses, and transition to fail-safe states during component failures. This research work advances safety protocols in human-robot collaboration within industrial settings, fostering a safe working environment through a well-defined safety framework. © 2025 IEEE.

关键词： Industrial robots

来源：评论

学校读者我要写书评

暂无评论

A Study on Efficiency Improvements of DNN Accelerators via Denormalized Arithmetic Elimination 38

A Study on Efficiency Improvements of DNN Accelerators via D...

引用

38th International Conference on VLSI Design, VLSID 2025

作者： Kharitonov, Alexander Kundu, Sandip Institute of Computer Architecture & Computer Engineering University of Stuttgart Stuttgart Germany University of Massachusetts Department of Electrical & Computer Engineering Amherst United States

ISBN: (纸本)9798331522445

Deep Neural Networks (DNNs) excel in tasks such as vision and speech processing, driving their widespread use. A key operation in DNNs is convolution, which is often implemented using General Matrix Multiplication (GEMM). Modern CPUs incorporate hardware accelerators for faster GEMM operation. Reduced precision floating-point (RPF) formats such as bfloat16 and TensorFloat32 are popular in machine learning, offering a balance between classification accuracy, dynamic range for training gradients, and hardware efficiency compared to traditional IEEE 754 32-bit or 64-bit floating-point numbers. While RPF is well-studied in the context of DNNs, a major complexity arises from handling denormalized numbers. This paper investigates their impact on DNNs and finds that removing denormalized number support from hardware accelerators does not affect classification accuracy, while allowing significant improvements in hardware efficiency and performance. With proliferation of data centers to accommodate the growing number of DNN applications, power reductions at the GEMM level can significantly improve cluster-wide efficiency, promoting sustainability. Our results indicate that eliminating denormalized arithmetic leads to significant improvements in multiplier circuit efficiency, with reductions of up to 46% in area (LUT count), 28% in longest path delay, and 59% in power consumption, all while maintaining DNN classification accuracy. © 2025 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Enhancing Privacy in IoT Networks: A Comparative Analysis of Classification and Defense Methods

引用

IEEE Access 2025年 13卷 71611-71646页

作者： Emre Ergun, Ahmet Can, Ozgu Kantarcioglu, Murat Ege University Graduate School of Natural and Applied Sciences Bornova35100 Turkey Izmir Katip Celebi University Faculty of Engineering and Architecture Department of Computer Engineering Cigli35620 Turkey Ege University Department of Computer Engineering Bornova35100 Turkey Virginia Tech Department of Computer Science BlacksburgVA24061 United States

The rapid proliferation of Internet of Things (IoT) devices has led to a substantial increase in network packet traffic, raising significant privacy concerns. Although traffic encryption is employed to protect the privacy of IoT devices, attackers can still leverage Machine Learning (ML) and Deep Learning (DL) techniques to classify device types by analyzing packet characteristics, such as size and timing. The main challenges in the state of the art are the lack of effective methods for exposing privacy violations in encrypted IoT traffic, and the absence of robust defense mechanisms to mitigate privacy breaches caused by network traffic analysis. Considering these challenges, this study presents two key contributions: (i) a novel vector-based classification method that enhances device-type identification from encrypted IoT traffic using advanced ML and DL techniques, and (ii) a robust defense mechanism based on Differential Privacy (DP) and advanced padding techniques against traffic analysis attacks. Therefore, the study examines privacy risks associated with sequential IoT device data and evaluates the effectiveness of ML algorithms using two datasets. The results demonstrate that the proposed vector-based classification method significantly improves the attacker’s classification accuracy, even when privacy-preserving techniques, such as padding, are used to obscure device-type classification. For this purpose, the study evaluates eXtreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) for IoT traffic classification, achieving an accuracy rate of 99.61% with XGBoost, 96.74% with LSTM, and 96.94% with GRU. Additionally, the Decision Tree (DT), Random Forest (RF), k-Nearest Neighbors (kNN), and GRU classification algorithms are also evaluated and compared with the XGBoost and LSTM classifiers for the proposed attack model. As a defense mechanism, DP is applied using the Fourier Perturbation Algorithm (FPA) to optimize padd

关键词： Decision trees

来源：评论

学校读者我要写书评

暂无评论

CNN-based cancer prediction scheme using 5G-assisted federated learning for healthcare Industry 5.0

引用

Alexandria Engineering Journal 2025年 126卷 131-142页

作者： Kumari, Aparna Patadia, Divya Tanwar, Sudeep Pau, Giovanni Alqahtani, Fayez Tolba, Amr Department of Computer Science and Engineering Institute of Technology Nirma University Ahmedabad382481 India Faculty of Engineering and Architecture Kore University of Enna Enna94100 Italy Software Engineering Department College of Computer and Information Sciences King Saud University Riyadh12372 Saudi Arabia Computer Science Department Community College King Saud University Riyadh11437 Saudi Arabia

Healthcare Industry 5.0 has witnessed a transformative change by incorporating Artificial Intelligence (AI), Internet-of-things (IoT), and genomics to provide individualized and patient-centered healthcare. It prioritizes comprehensive health, proactive strategies for chronic diseases like tumors, cancer, etc., and effective cooperation among participants. Much work has been done for cancer prediction using machine learning and deep learning approaches. However, it has yet to be fully explored to protect data privacy and security by limiting data sharing. So, this paper proposes a Convolutional Neural Networks (CNN)-based Cancer Prediction scheme, i.e., CNN-CPS within the context of Healthcare Industry 5.0. The proposed scheme harnesses the capabilities of 5G-assisted Federated Learning (FL), a revolutionary method for predicting cancer occurrences with data privacy, security, and limiting data sharing. Then, it also leverages model data from dispersed systems to boost efficiency. By capitalizing on the speed and efficiency of 5G networks and the collaborative nature of FL, precise cancer predictions can be attained, all while upholding the confidentiality of sensitive data. This integration of technologies has the potential to significantly reshape cancer diagnosis and therapeutic approaches in the Industry 5.0 landscape. Here, 5G can enable real-time data interchange across remote data centers with rapid transmission of data capabilities and low-latency connectivity, enabling more effective and precise model training. Hence, this study examines the FL to predict breast cancer and combines the optimizer weights from several clients to continually enhance the performance of the overall model. The experimental results show the efficacy of the CNN-CPS compared to the existing approaches based on several parameters like low latency (40.77% improvement), accuracy (75%), and loss (0.54). © 2025 The Authors

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Multimodal Input Classification for a Companion Pet Robot using Deep Learning 24

Multimodal Input Classification for a Companion Pet Robot us...

引用

8th International Conference on Advances in Artificial Intelligence, ICAAI 2024

作者： Alhouli, Sarah Yousef Almania, Nora Abdullah Sharma, Jitendra Kumar Sahoo, Deepak Ranjan Department of Computer Science Swansea University Swansea United Kingdom Shuwaikh Kuwait Department of Computer Science Shaqra University Shaqra Saudi Arabia Architecture & Town Planning Department Universite Libre de Bruxelles Bruxelles Belgium

ISBN: (纸本)9798400718014

Social robots have emerged recently as a technology that could provide social support and increase companionship, especially for older adults. Effective human-robot interaction requires social robots to interact empathically and naturally by using Artificial Emotional Intelligence (AEI), which allows them to recognise different multimodal inputs and engage with users emotionally. To this end, we developed AEI companion pet robot classification system that incorporated various multimodal inputs, including hand gestures, emotions from images, emotions from voice annotations, and speech detection. We utilised different deep learning approaches for each mode and adopted lightweight computationally efficient models. For multimodal input detection, we extended the MediaPipe framework with a mathematical approach for hand gestures, integrated an updated MobileNet transfer learning model for facial emotions, employed a Convolutional Neural Network (CNN) model for detecting emotions from voice, and used a lightweight Whisper model as an automatic speech recognition (ASR). Our findings indicated that the modified MobileNet lightweight model achieved an accuracy of 84.60% along with an efficient gesture detection system with an accuracy of 90% integrated with an effective CNN model of 86.89% accuracy. Our results imply that developing a multimodal AEI companion pet robot could promote future research in the development of emotionally intelligent companion robots. © 2024 Copyright held by the owner/author(s).

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Control of an Hybrid AC/DC MG Using Raspberry Pi

Control of an Hybrid AC/DC MG Using Raspberry Pi

引用

International Conference of the TC Electrimacs Committee, ELECTRIMACS 2024

作者： Carrasco-González, David Sarrias-Mena, Raúl Horrillo-Quintero, Pablo Llorens-Iborra, Francisco Fernández-Ramírez, Luis M. Cadiz Spain SURET Research Group Department of Engineering in Automation Electronics and Computer Architecture and Networks ETSI Algeciras Cadiz Spain

ISBN: (纸本)9783031739200

Nowadays, hybrid AC/DC microgrids (HMG) are growing in popularity because of human needs and their benefits. They are composed of the interconnection of a DC microgrid (MG) and an AC MG, and each MG is constituted by generation, energy storage, and consumption technologies. This study develops a new control strategy for an HMG consisting of a DC MG and an AC MG interconnected through a bidirectional power converter. The AC MG is composed of AC loads, a photovoltaic (PV) generator and a battery bank. The DC MG is made up of a hydrogen system, DC loads, wind turbine (WT) and ultracapacitor (UC). The HMG is evaluated under different conditions in an experimental setup constituted by an OPAL-RT 4512 unit and a Raspberry Pi microcontroller. This experimental setup permits to execute the HMG in real time and to validate the results obtained. Finally, the control system developed for this HMG presents a satisfactory behaviour under different working conditions. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Wind turbines

来源：评论

学校读者我要写书评

暂无评论

Reverse-Engineering Optimization Techniques of High-Level Synthesis: Practical Insights into Accelerating Applications with AMD-Xilinx Vitis

引用

IEEE Transactions on computer-Aided Design of Integrated Circuits and Systems 2025年

作者： Koronis, Jorge Garnica, Oscar Hidalgo, J. Ignacio Davila, Juan Lanchares Spain Department of Computer Architecture and Automation Faculty of Computer Science Universidad Complutense de Madrid Spain

Modern AI applications contain computationally expensive sections. Accelerator cards and tools like AMD Vitis HLS leverage high-level synthesis and hardware (HW) optimizations to create custom HW designs to accelerate ***, the learning curve is steep, even for those with previous knowledge of HW design, due to the complexity of the optimization techniques and limited information on their interactions and HW effects. This paper quantitatively analyzes the interactions of optimization techniques after reverse engineering Vitis' optimization directives, both in isolation and in *** 150 experiments were conducted to investigate three distinct goals: assessing pragma behavior and the rules governing pragma application and optimizations, modeling Vitis HLS latency estimates, and evaluating the impact of optimizations on design space exploration, specifically area and latency. These experiments involve different combinations and placements of optimizations in the loop and function hierarchy of the test bench. Our findings offer guidance on using Vitis pragmas and identify promising configurations for optimizing latency and area. © 1982-2012 IEEE.

关键词： Reverse engineering

来源：评论

学校读者我要写书评

暂无评论

A hypothetical comparative evaluationsystem for arctic indoors

引用

Frontiers of Architectural Research 2025年第1期14卷 210-223页

作者： Seyed-Amin Tabatabaeifard Jean-Francois Lalonde Marc Hebert Andre Potvin Claude MH.Demers GRAP(Groupe de Recherche en Ambiances Physiques) Ecole d’architectureUniversite LavalQuebecCanada Computer Vision and Systems Lab Department of Electrical and Computer EngineeringUniversite LavalQuebecCanada CERVO Brain Research Centre Faculty of MedicineUniversite LavalQuebecCanada

This research presents an innovative approach to evaluating indoor spaces,combining qualitative attributes with numerical architectural metrics. A hypothetical comparative visualization system is introduced, utilizing HDR visual imaging and thermal imaging in 360° field of view across multiple indoor environments. The study aims to provide architectsand occupants with a user-friendly tool informing them about the primary considerations oftheir built spaces, with a specific focus on indoor environmental qualities in remote Arctic regions. Key inquiries delve into the efficacy of the spherical approach and the capacity ofcomparative visualization to offer insights into space quality. Preliminary experiments contrastindoor environments in terms of circadian lighting, thermal uniformity, and view access tooutside in the 360° field of view (VAR360). The resulting visualizations hold significance inintroducing an immersive approach for depicting specific non-visible environmental qualities,particularly in relation to the window characteristics of spaces. It demonstrates the integration of multiple environmental variables, both steady-state and temporal, from central pointswithin spaces, providing a comprehensive view over their non-visible qualities. These resultsshould be useful for researchers and practitioners within building sciences, computer vision,and photobiology, showcasing an out-of-the-box approach for categorizing indoor spaces basedon standards and human-environmental qualifications.

关键词： Environmental assessment 360°imaging Comparative systems Low-cost tools Built environments

来源：评论

学校读者我要写书评

暂无评论

Machine learning model for predicting long-term energy consumption in buildings

引用

Discover Internet of Things 2025年第1期5卷 1-24页

作者： Hussien, Aseel Maksoud, Aref Al-Dahhan, Aisha Abdeen, Ahmed Baker, Thar Department of Architectural Engineering University of Sharjah Sharjah United Arab Emirates Department of Computer Science University of Sharjah Sharjah United Arab Emirates Civil Mining Environmental and Architectural Engineering University of Wollongong Wollongong Australia School of Architecture Technology and Engineering University of Brighton Brighton United Kingdom

The rapid growth in the construction sector has led to increased energy consumption and carbon emissions. Calculating energy usage and emissions is essential to energy security and promoting sustainable sector development. Therefore, the study objective is to investigate the utilazation of machine learning algorithm to predict long-term energy consumption in buildings sector, aiming to improve sustainable design and energy optimization, via the implementation of three machine learning models, XGBoost, Support Vector Regression, and Long-Short-Term Memory networks, to predict energy consumption. These models are adept at capturing complex interactions between building characteristics, environmental factors, and energy patterns. Although previous studies have explored various machine learning techniques for energy efficiency, limited research links these models to practical applications in building performance simulation. Furthermore, there is a lack of comparative evaluation of advanced machine learning models such as XGBoost, Support Vector Regression, and Long-Short-Term Memory to predict the energy consumption of building envelopes, particularly in hot climates such as the UAE. This research aims to fill this gap by providing a detailed comparison of these models against alternative approaches mentioned in the literature. The findings position Long-Short-Term Memory as a transformative force in predictive modeling, demonstrating exceptional precision with an R-squared value of 0.993 and a Mean Squared Error of 0.004. In contrast, Support Vector Regression and XGBoost showed limited predictive capabilities, with R-squared values of 0.462 and 0.94, respectively. This study establishes a solid data-driven foundation for architects and engineers to inform decisions on energy-efficient building designs, advocating Long-Short-Term Memory as the superior model for predicting energy performance. © The Author(s) 2025.

关键词： Green buildings

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：