检索结果-内蒙古大学图书馆

Domain generalization with semi-supervised learning for people-centric activity recognition

Science China(Information Sciences) 2025年第1期68卷 171-188页

作者： Jing LIU Wei ZHU Di LI Xing HU Liang SONG Academy for Engineering & Technology Fudan University Shanghai East-bund Research Institute on Networking Systems of AI School of Optoelectronic Information and Computer Engineering University of Shanghai for Science & Technology

People-centric activity recognition is one of the most critical technologies in a wide range of real-world applications,including intelligent transportation systems, healthcare services, and brain-computer interfaces. Large-scale data collection and annotation make the application of machine learning algorithms prohibitively expensive when adapting to new tasks. One way of circumventing this limitation is to train the model in a semi-supervised learning manner that utilizes a percentage of unlabeled data to reduce the labeling burden in prediction tasks. Despite their appeal, these models often assume that labeled and unlabeled data come from similar distributions, which leads to the domain shift problem caused by the presence of distribution gaps. To address these limitations, we propose herein a novel method for people-centric activity recognition,called domain generalization with semi-supervised learning(DGSSL), that effectively enhances the representation learning and domain alignment capabilities of a model. We first design a new autoregressive discriminator for adversarial training between unlabeled and labeled source domains, extracting domain-specific features to reduce the distribution gaps. Second, we introduce two reconstruction tasks to capture the task-specific features to avoid losing information related to representation learning while maintaining task-specific consistency. Finally, benefiting from the collaborative optimization of these two tasks, the model can accurately predict both the domain and category labels of the source domains for the classification task. We conduct extensive experiments on three real-world sensing datasets. The experimental results show that DGSSL surpasses the three state-of-the-art methods with better performance and generalization.

关键词： activity recognition deep learning domain generalization semi-supervised learning adversarial training

来源：评论

学校读者我要写书评

暂无评论

DTAIS:Distributed trusted active identity resolution systems for the Industrial Internet

引用

Digital Communications and Networks 2024年第4期10卷 853-862页

作者： Tao Huang Renchao Xie Yuzheng Ren F.Richard Yu Zhuang Zou Lu Han Yunjie Liu Demin Cheng Yinan Li Tian Liu State Key Lab.of Networking and Switching Technology Beijing University of Posts and TelecommunicationsBeijingChina Department of Systems and Computer Eng. Carleton UniversityOttawaONCanada

In recent years,the Industrial Internet and Industry 4.0 came into *** the development of modern industrial intelligent manufacturing technology,digital twins,Web3 and many other digital entity applications are also *** applications apply architectures such as distributed learning,resource sharing,and arithmetic trading,which make high demands on identity authentication,asset authentication,resource addressing,and service ***,an efficient,secure,and trustworthy Industrial Internet identity resolution system is ***,most of the traditional identity resolution systems follow DNS architecture or tree structure,which has the risk of a single point of failure and DDoS *** they cannot guarantee the security and privacy of digital identity,personal assets,and device *** we consider a decentralized approach for identity management,identity authentication,and asset *** this paper,we propose a distributed trusted active identity resolution system based on the inter-planetary file system(IPFS)and non-fungible token(NFT),which can provide distributed identity resolution *** we have designed the system architecture,identity service process,load balancing strategy and smart contract *** addition,we use Jmeter to verify the performance of the system,and the results show that the system has good high concurrent performance and robustness.

关键词： Industrial Internet NFT IPFS Trust Identity resolution system

来源：评论

学校读者我要写书评

暂无评论

Intelligent Detection System for a Distributed Denial-of - Service (DDoS) Attack Based on Time Series 15

Intelligent Detection System for a Distributed Denial-of - S...

引用

15th International Conference on Developments in esystems Engineering, DeSE 2023

作者： Alsumaidaie, Mustafa S. Ibrahim Alheeti, Khattab M. Ali Al-Aloosy, Abdul Kareem University of Anbar Department of Computer Science Ramadi Iraq University of Anbar Computer Networking Systems Department Ramadi Iraq

ISBN: (纸本)9798350335149

With a surge in the usage of systems that largely depend on networking and programming, the need for cybersecurity has grown as well. Cyberattacks are a rising threat to companies and people. The Distributed Denial of Service (DDoS) attack is one of the destructive hacks that have swiftly acquired appeal among hackers. In this work, a security system is proposed to prevent DDoS. In other words, it has the ability to protect external and internal communication systems from attacks. The primary contribution of this work is to acquire the best accuracy based on time series. Multiple machine learning algorithms are applied and compared between them. The Random Forest accuracy is 100% and the XGBoost was 91% using the same data set. © 2023 IEEE.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

A Mobile-Based Application for Enhance the Knowledge of Kindergarten Children 5

A Mobile-Based Application for Enhance the Knowledge of Kind...

引用

5th International Conference on Advancements in Computing, ICAC 2023

作者： Sathintha, Tharuka Basnayake, Omal Wickramasinghe, Piyumi Dharmakeerthi, Rusiru Dharmakeerthi, Uditha Jayakody, Anuradha Sri Lanka Institute of Information Technology Department of Computer Systems Networking Malabe Sri Lanka

ISBN: (纸本)9798350358131

This proposal outlines the development of a mobile application aimed at enhancing the learning capabilities of Sri Lankan kindergarteners. Through the improvement of early writing and speaking skills, emotion recognition, and personalized memory assessments, the application seeks to identify and address learning gaps in young children. The application's core focus includes four key components: refining writing skills through guided practice, empowering parents to identify and track their child's weaknesses and progress, enabling accurate pronunciation practice and tracking, and promoting emotional awareness by identifying and reflecting emotions through music. Additionally, the app offers tailored memory assessments via a mobile game, catering to individual learning needs and fostering cognitive growth. Extensive research, collaboration with educators, psychologists, and parents, as well as usability testing, contribute to the app's effectiveness and user-friendliness. By providing personalized learning experiences, the mobile application has the potential to significantly enhance educational outcomes for Sri Lankan children, equipping them with essential skills in writing, speaking, emotional intelligence, and memory retention. © 2023 IEEE.

关键词： E-learning

来源：评论

学校读者我要写书评

暂无评论

Reduced Spread Risk of COVID-19 based on Tracking System Monitoring in WSNs 16

Reduced Spread Risk of COVID-19 based on Tracking System Mon...

引用

16th International Conference on Developments in esystems Engineering, DeSE 2023

作者： Fahed, Majed H. Rashid, Ahmed N. University of Anbar Computer Sciences College of Computer and Information Technology Iraq University of Anbar Computer Networking Systems College of Computer and Information Technology Iraq

ISBN: (纸本)9798350381344

The phrase 'COVID-19' has become the most popular and highly searched term on Google since its emergence in November 2019. The world must combat this pandemic. With advancements in mobile technology and sensors, Developing Internet of Things (IoT)-based healthcare solutions is now feasible. Proactive as well as preventative, these new systems, are in contrast to the traditional reactive healthcare systems. In this study, a paradigm based on the Internet of Things is proposed for detecting COVID-19 suspects. The framework presented in this research focuses on predicting COVID-19 in its early stages. To determine whether the virus is present, it uses health measurements collected from sensors and other Internet of Things devices. By gathering symptom data related to COVID-19, the behavior of the virus can be better understood. Four machine learning models are introduced in this paper: CatBoost, Decision Tree, Random Forest, and Logistic Regression. These models are used to create a Diagnostic Learning Classifier (DLC) model. The top three performing models, Decision Tree, Logistic Regression, and CatBoost, are combined using a weighted average or majority voting approach, depending on the nature of the problem. The DLC model is compared to base models such as Random Forest, Decision Tree, Logistic Regression, and CatBoost, using multiple evaluation criteria. The goal of this ensemble model is to increase forecast accuracy by utilizing the variety and advantages of the chosen models. It delivers great diagnostic accuracy when used as a machine learning model on the gathered data. Notably, the DLC classifier achieves an accuracy of 91%, followed by Random Forest (86%), CatBoost (85%), Logistic Regression (84%), and Decision Tree (82%). These results demonstrate the effectiveness of the voting classifier in efficiently managing COVID-19 by accurately diagnosing patients and notifying health authorities of suspected cases. © 2023 IEEE.

关键词： COVID-19

来源：评论

学校读者我要写书评

暂无评论

Dynamic and Private Recommendation System

Dynamic and Private Recommendation System

引用

2023 International Conference on Digital Applications, Transformation and Economy, ICDATE 2023

作者： Lo, Gee Yen Tan, Terence Peng Lian Curtin University Malaysia Computer Systems and Networking Miri Malaysia Curtin University Malaysia Department of Electrical & Computer Engineering Miri Malaysia

ISBN: (纸本)9798350310689

The amount of data that we retrieve and use every day is increasing rapidly. As data naturally exist in different data domains, the number of data domains that we retrieve and use every day has also increased as a result. As different recommendation systems are needed for different data domains, more and more recommendation systems will need to be built. Furthermore, users will lose their privacy as many recommendation systems collect information about their users. This paper aims to solve the aforementioned issues by creating a recommendation system that will work with any data domain and upholds the privacy of its users. The Dynamic Implementation Of Privacy Based Recommendation System Application is created in this paper. The objective of this thesis is achieved as the Dynamic Implementation Of Privacy Based Recommendation System Application developed in this paper upholds the privacy of its users as it only stores users' data in their device's local database and the application has the ability to adapt to different data domains. © 2023 IEEE.

关键词： Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Lung Cancer Detection Using Machine Learning and Deep Learning Models 21

Lung Cancer Detection Using Machine Learning and Deep Learni...

引用

21st International Multi-Conference on systems, Signals and Devices, SSD 2024

作者： Alheeti, Khattab M. Ali Al-Shouka, Tabreer T. Majeed, Saleem Hamad Ahmed, Amjed Abbas College of Computer and Information Technology University of Anbar Computer Networking Systems Department Anbar Iraq College of Computer and Information Technology University of Anbar Computer Sciences Department Anbar Iraq Department of Computer Techniques Engineering Baghdad Iraq

ISBN: (纸本)9798350374131

Technology and artificial intelligence play a significant role in improving healthcare and enable tasks to be automated. In addition, the diseases can be better understood and diagnosed faster, saving time and reducing costs. This study examines the impact of transfer learning models on the effectiveness of deep learning models in classifying lung cancer through the analysis of CT scan images. Additionally, it investigates the relative performance of various machine learning and deep learning models, encompassing Support Vector Machine (SVM) and convolutional neural net-works (CNN) such as InceptionV3, VGG16, Xception, ResNet50, and MobileNetV2, in the early detection of lung cancer based on CT scan images. The SVM model achieved an overall accuracy of 89% after preprocessing, the proposed approach was applied to five pre-trained models (ResNet50, In-ceptionV3, VGG16, Xception, MobileNetV2) using the dataset: Chest CT-Scan;Among the pre-trained CNN models, the Mo-bileNetV2 model achieved the highest accuracy of 98% and the lowest test loss, indicating it performed the best. The Xception model achieved the second-highest accuracy of 97%. The image pre-processing phase plays a significant role in im-proving system performance in terms of improving image contrast and increasing processing speed. © 2024 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

BitMoD: Bit-serial Mixture-of-Datatype LLM Acceleration 31

BitMoD: Bit-serial Mixture-of-Datatype LLM Acceleration

引用

31st IEEE International Symposium on High Performance computer Architecture, HPCA 2025

作者： Chen, Yuzong Abouelhamayed, Ahmed F. Dai, Xilai Wang, Yang Andronic, Marta Constantinides, George A. Abdelfattah, Mohamed S. Cornell University Computer Systems Lab United States Systems and Networking Research Group Microsoft Research United States Imperial College London Department of Electrical and Electronic Engineering United Kingdom

ISBN: (纸本)9798331506476

Large language models (LLMs) have demonstrated remarkable performance across various machine learning tasks. Yet the substantial memory footprint of LLMs significantly hinders their deployment. In this paper, we improve the accessibility of LLMs through BitMoD1, an algorithm-hardware co-design solution that enables efficient LLM acceleration at low weight precision. On the algorithm side, BitMoD introduces fine-grained data type adaptation that uses a different numerical data type to quantize a group of (e.g., 128) weights. Through the careful design of these new data types, BitMoD is able to quantize LLM weights to very low precision (e.g., 4 bits and 3 bits) while maintaining high accuracy. On the hardware side, BitMoD employs a bitserial processing element to easily support multiple numerical precisions and data types;our hardware design includes two key innovations: First, it employs a unified representation to process different weight data types, thus reducing the hardware cost. Second, it adopts a bit-serial dequantization unit to rescale the per-group partial sum with minimal hardware overhead. Our evaluation on six representative LLMs demonstrates that BitMoD significantly outperforms state-of-the-art LLM quantization and acceleration methods. For discriminative tasks, BitMoD can quantize LLM weights to 4 -bit with © 2025 IEEE.

关键词： Integrated circuit design

来源：评论

学校读者我要写书评

暂无评论

A Transfer Learning for Intelligent Prediction of Lung Cancer Detection

A Transfer Learning for Intelligent Prediction of Lung Cance...

引用

2023 Al-Sadiq International Conference on Communication and Information Technology, AICCIT 2023

作者： Al-Shouka, Tabreer T. Alheeti, Khattab M. Ali University of Anbar College of Computer and Information Technology Department of Computer Sciences Anbar Iraq Technology University of Anbar College of Computer and Information Department of Computer Networking Systems Iraq

ISBN: (纸本)9798350341881

Cancer is a widespread and potentially lethal illness. There are many different types of cancer. The most common type of cancer is lung cancer. High mortality rates from lung cancer are major public health concerns. Therefore, early detection of cancer reduces the possibility of dying from the disease. A CT scan is used to determine the best course of treatment. Screening with CT has been shown to reduce mortality from lung cancer by detecting the disease at its earliest stages when it manifests as pulmonary nodules. Artificial Intelligence (AI) systems, which employ transfer-learning models, have shown great potential in enhancing the precision and speed of lung cancer diagnosis and treatment. Transfer learning enables adapting pre-existing models to new tasks, facilitating more efficient and effective medical data analysis. Additionally, these systems can assist clinicians in identifying the specific type and stage of lung cancer, which can guide treatment decisions. Our study explores the use of CNN transfer learning with RESNET, MobileNetV2, Xception, and VGG16 models for medical image analysis tasks;res-net achieved the highest testing accuracy in our research, at 0.94, with a testing loss of0.16, which has demonstrated the potential to enhance the accuracy and efficiency of healthcare AI systems. © 2023 IEEE.

关键词： Health care

来源：评论

学校读者我要写书评

暂无评论

ISAC-Driven Channel-State-Estimation for Statistical-QoS Based MURLLC Over 6G M-MIMO Mobile Networks Using FBC

ISAC-Driven Channel-State-Estimation for Statistical-QoS Bas...

引用

2024 IEEE Global Communications Conference, GLOBECOM 2024

作者： Zhang, Xi Zhu, Qixuan Texas A&m University Networking and Information Systems Laboratory Department of Electrical and Computer Engineering College StationTX77843 United States

ISBN: (纸本)9798350351255

The massive ultra-reliable and low-latency communications (mURLLC) services are emerging as a new traffic type to support communications among massive numbers of mobile users (MUs) demanding the stringent delay and error-rate bounded quality-of-services (QoS) requirements in 6G. Correspondingly, the integrated sensing and communications (ISAC) technique can improve various QoS performances through spectrum/hardware sharing between target-sensing and data-communications. In addition, the massive multiple-input multiple-output (massive MIMO) and finite blocklength coding (FBC) have been also widely recognized as the powerful 6G techniques to efficiently support mURLLC. However, how to apply the ISAC technique to accurately sense and estimate the wireless channel state while transmitting the information from the base station to MUs to support mURLLC using massive MIMO and FBC has imposed many new difficulties not encountered before. To conquer these challenges, in this paper we propose the new ISAC performance metrics for statistical-QoS guaranteed mURLLC over 6G massive MIMO mobile networks using FBC. First, we establish system models for ISAC-based architectures. Second, we detect targeted MUs and estimate the channel state using the radar sensing echo. Third, we develop a new ISAC performance metric as a function of the data-communication rate and channel state estimation distortion. Finally, we use the numerical analyses to validate and evaluate our developed ISAC performances in supporting mURLLC-oriented QoS requirements. © 2024 IEEE.

关键词： Channel estimation

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：