检索结果-内蒙古大学图书馆

IEEE International Conference on Communications (ICC)

作者： Shisheng Hu Jie Gao Xinyu Huang Mushu Li Kaige Qu Conghao Zhou Xuemin Shen Department of Electrical and Computer Engineering University of Waterloo Canada School of Information Technology Carleton University Canada Department of Electrical Computer and Biomedical Engineering Toronto Metropolitan University Canada

ISBN: (数字)9781728190549

ISBN: (纸本)9781728190556

In this paper, we propose a digital twin (DT)-based user-centric approach for processing sensing data in an integrated sensing and communication (ISAC) system. The considered scenario involves an ISAC device with a lightweight deep neural network (DNN) and a mobile edge computing (MEC) server with a large DNN. After collecting sensing data, the ISAC device either processes the data locally or uploads them to the server for higher-accuracy data processing. To cope with data drifts, the server updates the lightweight DNN when necessary, referred to as continual learning. Our objective is to minimize the long-term average computation cost of the MEC server by jointly optimizing two decisions, i.e., sensing data offloading and sensing data selection for the DNN update. A DT of the ISAC device is constructed to predict the impact of potential decisions on the long-term computation cost of the server, based on which the decisions are made with closed-form formulas. Experiments on executing DNN-based human motion recognition tasks are conducted to demonstrate the outstanding performance of the proposed DT-based approach in computation cost minimization.

关键词： Continuing education Performance evaluation Costs Multi-access edge computing Training data Artificial neural networks Integrated sensing and communication

来源：评论

学校读者我要写书评

暂无评论

XBreaking: Explainable Artificial Intelligence for Jailbreaking LLMs

arXiv

引用

arXiv 2025年

作者： Arazzi, Marco Kembu, Vignesh Kumar Nocera, Antonino Vinod, P. Department of Electrical Computer and Biomedical Engineering University of Pavia Italy Department of Computer Applications Cochin University of Science & Technology India

Large Language Models are fundamental actors in the modern IT landscape dominated by AI solutions. However, security threats associated with them might prevent their reliable adoption in critical application scenarios such as government organizations and medical institutions. For this reason, commercial LLMs typically undergo a sophisticated censoring mechanism to eliminate any harmful output they could possibly produce. In response to this, LLM Jailbreaking is a significant threat to such protections, and many previous approaches have already demonstrated its effectiveness across diverse domains. Existing jailbreak proposals mostly adopt a generate-and-test strategy to craft malicious input. To improve the comprehension of censoring mechanisms and design a targeted jailbreak attack, we propose an Explainable-AI solution that comparatively analyzes the behavior of censored and uncensored models to derive unique exploitable alignment patterns. Then, we propose XBreaking, a novel jailbreak attack that exploits these unique patterns to break the security constraints of LLMs by targeted noise injection. Our thorough experimental campaign returns important insights about the censoring mechanisms and demonstrates the effectiveness and performance of our attack. © 2025, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Dual Dynamic Consistency Regularization for Semi-Supervised Domain Adaptation

引用

IEEE Access 2024年 12卷 36267-36279页

作者： Ngo, Ba Hung Lam, Ba Thinh Nguyen, Thanh Huy Dinh, Quang Vinh Choi, Tae Jong Chonnam National University Graduate School of Data Science Gwangju61186 Korea Republic of University of Science Faculty of Information Technology Ho Chi Minh City700000 Viet Nam National Cheng Kung University Department of Biomedical Engineering Tainan70101 Taiwan Vietnamese-German University School of Electrical Engineering and Computer Science Ho Chi Minh City72000 Viet Nam

The Vision Transformer (ViT) model serves as a powerful model to capture and comprehend global information, particularly when trained on extensive datasets. Conversely, the Convolutional Neural Network (CNN) model is beneficial to training with small datasets for retaining essential local information. Inspired by these properties of ViT and CNN models, we introduce a hybrid framework that smoothly increases the cross-domain generalization in Semi-supervised Domain Adaptation (SSDA). To achieve this goal, we first train the ViT model on abundant labeled source data, while the CNN model is trained on a few labeled target samples. Then, these models dynamically exchange their knowledge for potential generalization to unlabeled target data via the proposed method, named Dual Dynamic Consistency Regularization (D2CR). Specifically, the ViT model provides its pseudo labels to update the global perspective for the CNN model. Similarly, the CNN model offers pseudo labels to complement the local perspective for the ViT model. The previous methods use a fixed threshold algorithm for the pseudo-labeling process. However, we utilize the dynamic threshold strategy to create pseudo labels for the bi-directional consistency regularization learning between the ViT and CNN models. We verify our approach across several SSDA benchmark datasets. The outstanding experimental results provide strong evidence of the effectiveness and superiority of our approach over previous state-of-the-art SSDA methods. © 2013 IEEE.

关键词： Neural network models

来源：评论

学校读者我要写书评

暂无评论

Safety Comparison of Vehicular Platoons Under Different Bidirectional Communication Topologies: A Signal Analysis Method

Safety Comparison of Vehicular Platoons Under Different Bidi...

引用

International Conference on Intelligent Transportation

作者： Amir Zakerimanesh Tony Z. Qiu Mahdi Tavakoli Department of Electrical and Computer Engineering University of Alberta Edmonton AB Canada Department of Civil and Environmental Engineering University of Alberta Edmonton AB Canada Department of Electrical and Computer Engineering and the Department of Biomedical Engineering University of Alberta Edmonton AB Canada

ISBN: (数字)9798331505929

ISBN: (纸本)9798331505936

This paper conducts a comparative analysis of bidirectional communication topologies in vehicular platooning, emphasizing their impact on safety during travel. Introducing two novel metrics, the Accumulative Average Penalty of Minimum Time to Collision (AAPMTTC) and Accumulative Average Deceleration Rate to Avoid Collision (AADRAC), the study evaluates collision susceptibility between neighboring vehicles, taking into consideration their relative velocities and accelerations. Utilizing platoon dynamics based on intervehicle distances and their derivatives, the analysis captures the evolving vehicle behaviors over travel time. Results highlight the safety benefits of communication topologies where follower vehicles receive information from more preceding vehicles, particularly when incorporating the leader vehicle's state. This research underscores the significance of information exchange within vehicular platoons and offers insights for providing more safety through design of communication structure in automated driving scenarios in vehicular platooning.

关键词： Measurement Laplace equations Bidirectional control Safety Topology Vehicle dynamics Signal analysis Standards Intelligent transportation systems Information exchange

来源：评论

学校读者我要写书评

暂无评论

Adaptive Thermal Clothing with IoT and Random Forest Regression for Dynamic Outdoor Comfort

Adaptive Thermal Clothing with IoT and Random Forest Regress...

引用

2024 International Conference on Advances in Modern Age Technologies for Health and engineering Science, AMATHE 2024

作者： Srinivasan, S. Sridevi, V. Saravanan, K. Murugan, S. Srinivasan, C. Muthulekshmi, M. Saveetha School Of Engineering Saveetha Institute Of Medical And Technical Sciences Saveetha University Department Of Biomedical Engineering Tamil Nadu Chennai India Academy Of Maritime Education And Training Deemed To Be University Department Of Electrical And Electronics Engineering Tamil Nadu Chennai India R.M.D. Engineering College Department Of Information Technology Tamil Nadu Chennai India Saveetha School Of Engineering Saveetha Institute Of Medical And Technical Sciences Saveetha University Department Of Computer Science And Engineering Tamil Nadu Chennai India

ISBN: (纸本)9798350371567

This research presents a novel method for improving outdoor comfort by combining adaptive thermal apparel with Internet of Things (IoT) and Random Forest Regression. Wearing inflexible traditional outdoor gear may be a real pain for those living in different parts of the world. As a solution, we present an intelligent garment system that can adapt its thermal insulation in real-time using data gathered from IoT sensors. A Random Forest Regression model is used to ascertain the most appropriate thermal settings for the garments based on the gathered data, which encompasses environmental factors such as humidity, temperature, and wind speed. By taking each user's tastes and physiological needs into account, the adaptive system strives to provide individualized comfort. The suggested technique successfully maintains the target comfort levels throughout varied outside situations, according to the experimental findings. In an age of rising climatic fluctuation, our study offers a potential answer for dynamic outdoor comfort to the expanding area of smart fabrics and wearable technologies. © 2024 IEEE.

关键词： Energy efficiency

来源：评论

学校读者我要写书评

暂无评论

Prediction of Post-COVID-19 Pandemic Learning Mode Preferences Using Machine Learning 14

Prediction of Post-COVID-19 Pandemic Learning Mode Preferenc...

引用

14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023

作者： Pratondo, Agus Novianty, Astri Putro, Budi Laksono Komala Sari, Siska Telkom University School of Applied Sciences Department of Multimedia Engineering Bandung Indonesia Telkom University School of Electrical Engineering Department of Computer Engineering Bandung Indonesia Indonesia University of Education Department of Computer Science Education Bandung Indonesia Telkom University School of Applied System Diploma Program of Information System Bandung Indonesia

ISBN: (纸本)9798350335095

The COVID-19 epidemic has had a huge impact on the educational landscape, prompting the adoption of online and remote learning as viable alternatives to conventional in-person instruction. In order to create effective educational methods as institutions prepare for the post-pandemic period, it is essential to understand students' preferences for various learning modes. With the aid of machine learning techniques, this study seeks to forecast preferences for learning styles in post-pandemic higher education. The study made use of a dataset compiled from a wide range of college students, which included information on the institution type (public or private), program type (bachelor's or diploma), academic discipline (science or humanities), gender, year of study, internet use, communication costs, comorbidity status, vaccination history, gadget and laptop ownership, and preferred learning mode (online, in-person, or both). For the aim of prediction, a Support Vector Machine (SVM) was used. Its kernel functions included linear, polynomial, radial basis function (RBF), and sigmoid. Performance criteria like accuracy, precision, recall, and F1 score were used to assess the models after the dataset was randomly partitioned into training and testing sets. The findings showed that for identifying the preferred mode of learning, SVM with the RBF kernel had the highest prediction accuracy (91.67%). Further proving the RBF kernel's efficacy, the confusion matrix analysis showed that high values in the diagonal represented accurate predictions for each of the three classes (both online and onsite). The results imply that machine learning approaches, in particular SVM with the RBF kernel, may accurately forecast students' preferences for learning modes in the post-pandemic situation. In a post-pandemic society, the findings provide insightful information that may be used to build adaptable and inclusive learning models and put them into practice. © 2023 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

A Compact-Area Low-Power Temperature Sensor Featuring High Supply Voltage Scalability

A Compact-Area Low-Power Temperature Sensor Featuring High S...

引用

IEEE International Conference on Electronics, Circuits and Systems (ICECS)

作者： Elisabetta Moisello Piero Malcovati Edoardo Bonizzoni Department of Electrical Computer and Biomedical Engineering University of Pavia Italy

This paper proposes a highly supply voltage-scalable, low-power and compact-area temperature sensor, which satisfies the requirements for Systems-on-Chip (SoCs) and microprocessors hotspots monitoring. The proposed novel sensor structure employs diode-connected NMOS devices operating in sub-thresold, featuring different sizes, as sensing elements, in order to implement a proportional-to-absolute-temperature (PTAT) sensor output voltage. The proposed sensor does not require any additional biasing or start-up circuits and, since it mostly processes the signals generated by the sensing elements in the current domain, enables high supply voltage scalability. Furthermore, a low-cost 1-point temperature calibration algorithm can be employed for trimming the sensor sensitivity. The proposed sensor circuit was designed in a 130-nm CMOS process and its performance was verified through extensive simulations in the Cadence Virtuoso environment.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A D-Band Low-Noise-Amplifier in SiGe BiCMOS with Broadband Multi-Resonance Matching Networks

A D-Band Low-Noise-Amplifier in SiGe BiCMOS with Broadband M...

引用

European (Formerly European Gallium Arsenide and Other Semiconductors Application Symposium - EGASS) Microwave Integrated Circuit Conference (EuMIC)

作者： Guglielmo De Filippi Lorenzo Piotto Andrea Bilato Andrea Mazzanti Department of Electrical Computer and Biomedical Engineering University of Pavia Italy

Silicon amplifiers in D-Band are required to operate at high gain-bandwidth products and close to the cutoff frequency $f_{\max }$. Multi-stage amplifiers commonly employ stagger-tuning to meet the desired bandwidth, but with sub-optimal noise and linearity. Better performance is achieved with broadband inter-stage matching and gain progressively distributed among the stages. This work proposes a design flow for broadband matching networks approximating the response of a doubly-tuned transformer. The technique is applied to design a 3-stage D-band LNA in BiCMOS 55 nm technology. Measurements show 28 dB gain, 127-168 GHz bandwidth, NF down to 5.2 dB and >2dBm output compression point with 30 mA DC current from 2V supply. The performance compare favorably against previous works.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Introducing Spectral Attention for Long-Range Dependency in Time Series Forecasting 38

Introducing Spectral Attention for Long-Range Dependency in ...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Kang, Bong Gyun Lee, Dongjun Kim, HyunGi Chung, DoHyun Yoon, Sungroh Interdisciplinary Program in Artificial Intelligence Seoul National University Korea Republic of Department of Electrical and Computer Engineering Seoul National University Korea Republic of Department of Future Automotive Mobility Seoul National University Korea Republic of

Sequence modeling faces challenges in capturing long-range dependencies across diverse tasks. Recent linear and transformer-based forecasters have shown superior performance in time series forecasting. However, they are constrained by their inherent inability to effectively address long-range dependencies in time series data, primarily due to using fixed-size inputs for prediction. Furthermore, they typically sacrifice essential temporal correlation among consecutive training samples by shuffling them into mini-batches. To overcome these limitations, we introduce a fast and effective Spectral Attention mechanism, which preserves temporal correlations among samples and facilitates the handling of long-range information while maintaining the base model structure. Spectral Attention preserves long-period trends through a low-pass filter and facilitates gradient to flow between samples. Spectral Attention can be seamlessly integrated into most sequence models, allowing models with fixed-sized look-back windows to capture long-range dependencies over thousands of steps. Through extensive experiments on 11 real-world time series datasets using 7 recent forecasting models, we consistently demonstrate the efficacy of our Spectral Attention mechanism, achieving state-of-the-art results. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Two-step registration of rigid and non-rigid MR-iUS for brain shift compensation using transfer learning

Two-step registration of rigid and non-rigid MR-iUS for brai...

引用

International Symposium on Artificial Intelligence and Signal Processing (AISP)

作者： Amir Hossein Jalalzadeh Seyed Saeid Talebi Mehrdad Hashemi Kamangar Department of Biomedical Engineering Engineering faculty Shahed University Tehran Iran Department of Electrical and Computer Engineering Noshirvani University Babol Iran Department of Electrical and Electronic Engineering Engineering Faculty Shomal University Amol Iran

Accurate removal of brain tumors is always one of the most important challenges for surgeons, as the continuous change of the brain state after opening the skull and releasing the resulting pressure causes the tumor state to change. By registration of preoperative MR images on intraoperative ultrasound images, the extent of this change is estimated and a new image of the brain is created. The result shows the changes and shifts in the brain, and the surgeon removes the tumor based on this image. Image registration using new deep learning methods has attracted the attention of many researchers due to its high efficiency and accuracy. In this paper, the images of 22 male and female patients with grade 2 glioma tumors were used to evaluate the proposed method. MR images of the patients were taken before surgery, while ultrasound images were taken during surgery and after cranial incision. The deep network used in this paper to compensate for non-rigid changes is voxel morph. All images were fed to the pre-trained network in pairs, and the results are reported for each individual. The average error of all images for the proposed method is 3.56 ± 1.72. This shows the improvement in performance compared to the previous methods since in this work the landmarks were not used in training phase.

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：