检索结果-内蒙古大学图书馆

A novel approach to deriving adaboost classifier weights using squared loss function for overlapping speech detection

引用

Multimedia Tools and Applications 2025年 1-28页

作者： Asbai, Nassim Bounazou, Hadjer Zitouni, Sihem Department of Electronics Faculty of Sciences and Technology Mohamed El Bachir El Ibrahimi University Bordj Bou Arreridj City34030 Algeria Speech Com. & Signal Proc. Lab. Faculty of Electrical Engineering USTHB Bab Ezzouar City16111 Algeria Department of Computer Science Faculty of Mathematics and Computer Science Mohamed El Bachir El Ibrahimi University Bordj Bou Arreridj City34030 Algeria

In this paper, we address the complex problem of detecting overlapping speech segments, a key challenge in speech processing with applications in speaker diarization, automatic transcription, and multi-speaker recognition systems. Traditional approaches, often relying on exponential loss functions within the AdaBoost framework, struggle to maintain robustness in noisy or imbalanced data environments. To enhance detection accuracy, we propose a novel AdaBoost variant—SLFAdaBoost—utilizing a squared loss function specifically tailored for overlapping speech detection. This model demonstrates superior stability and convergence, significantly outperforming conventional methods. Experimental evaluation on the NIST 2005 corpus revealed that SLFAdaBoost achieves an accuracy range of 92.5% to 98.6%, with F1 scores between 95.4% and 99.2%, notably surpassing standard AdaBoost and Random Forest classifiers. Additionally, our model shows resilience in noisy conditions, maintaining precision and recall metrics at higher noise levels than other classifiers. These results underscore SLFAdaBoost's capacity to handle intricate data patterns, offering a more robust solution for real-world speech processing applications where overlapping segments are prevalent. This contribution provides an efficient, high-performing model, advancing the capabilities of ensemble methods in complex audio environments. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

Early Detection of Childhood Malnutrition Using Survey Data and Machine Learning Approaches 4

Early Detection of Childhood Malnutrition Using Survey Data ...

引用

4th International Conference on Sentiment Analysis and Deep Learning, ICSADL 2025

作者： Zumma, Thoufiq Rahaman, Anikur Islam Prova, Nuzhat Noor Haque, Tasriful Joy Bose, Julias Caeser Youki, Ravina Akter Daffodil International University Department of Computer Science and Engineering Dhaka Bangladesh Washington University Science and Technology Department of Computer Science New York United States Pace University Department of Seidenberg School of Csis New York United States Victorian Institute of Technology Master of Information Technology and Systems Australia University of Wollongong Faculty of Artificial Intelligence and Science Australia University of Technology Sydney Faculty of Engineering and It Australia

ISBN: (纸本)9798331523923

Children's physical, mental and emotional development depends heavily on sleep, with age-specific sleep needs fluctuating. Malnutrition may result from eating too little, absorbing nutrients poorly, being unwell, living in poverty or not knowing enough about nutrition. Food allergies and iron insufficiency are two examples of illnesses whose onsets might differ depending on environmental exposures and genetics. By providing proper diet, getting medical attention when necessary and encouraging healthy behaviors, parents play a crucial role in supporting their children's health. For children to grow and develop properly, they need meals high in fruits, vegetables, milk, protein, calcium, iron and fiber. Key advantages of a balanced diet include aiding bone formation, improving heart health and encouraging regular bowel motions and digestion. In general, children's general health and wellness depend on caregivers being aware of their dietary requirements and giving them the proper attention. The findings of the research imply that machine learning algorithms may be useful in understanding and predicting the prevalence of malnutrition in children. Furthermore, these algorithms could provide perceptive data that might help politicians create effective policies. In this research, we make use of five different machine learning algorithms. With a decision tree, the accuracy is 0.92%, followed by svm which provides 0.91%, KNN, which provides 0.84%, GBM classifier, which provides 0.79% and random forest, which provides 0.65%. © 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

LayerCFL:an efcient federated learning with layer-wised clustering

引用

Cybersecurity 2025年第1期8卷 72-85页

作者： Jie Yuan Rui Qian Tingting Yuan Mingliang Sun Jirui Li Xiaoyong Li School of Cyberspace Security Beijing University of Posts and TelecommunicationsBeijingChina Key Laboratory of Trustworthy Distributed Computing and Service(BUPT) Ministry of EducationBeijingChina Institute of Computer Science Faculty of Mathematics and Computer ScienceUniversity of GoettingenGöettingenGermany School of Information Technology Henan University of Chinese MedicineZhengzhouHenanChina

Federated Learning(FL)sufers from the Non-IID problem in practice,which poses a challenge for efcient and accurate model *** address this challenge,prior research has introduced clustered FL(CFL),which involves clustering clients and training them *** its potential benefts,CFL can be computationally and communicationally expensive when the data distribution is unknown *** is because CFL involves the entire neural networks of involved clients in computing the clusters during training,which can become increasingly timeconsuming with large-sized *** tackle this issue,this paper proposes an efcient CFL approach called LayerCFL that employs a Layer-wised clustering *** LayerCFL,clients are clustered based on a limited number of layers of neural networks that are pre-selected using statistical and experimental *** experimental results demonstrate the efectiveness of LayerCFL in mitigating the impact of Non-IID data,improving the accuracy of clustering,and enhancing computational efciency.

关键词： Federated learning Clustered federated learning Non-IID Layer-wised clustering

来源：评论

学校读者我要写书评

暂无评论

Are Fact Checkers Effective in the Post Truth World? Assessing Impact of Fact Checkers Cross Medium and Platforms 25th

Are Fact Checkers Effective in the Post Truth World? Asses...

引用

25th International Conference on Web Information Systems Engineering, WISE 2024

作者： Masurkar, Hrishikesh Suleiman, Basem Yaqub, Waheeb Alibasa, Muhammad Johan School of Computer Science The University of Sydney Sydney Australia School of Computer Science and Engineering The University of New South Wales Sydney Australia Faculty of Engineering and Technology Sampoerna University Jakarta Indonesia

ISBN: (纸本)9789819605668

Social media platforms help users share opinions and find new information but also spread rumors, which misinforms the public. These rumour threads often prompt users (called guardians) to respond with fact-checking articles to debunk or verify the rumour. We examine how general users’ and rumour starter respond to these fact-checking (FC) comments. This study explores these aspects by analysing users’ textual and non-textual engagements. These engagements are examined across three main factors: platforms (Twitter and Reddit), rumour topics, and the primary medium (text, image, video, and GIF) in the rumour post. We observed minimal differences across primary mediums by analyzing 147K news posts and FC comments with 20.7 million replies. Furthermore, we found that Reddit users engage with FC comments more than Twitter users in terms of likes and the rumour starter’s reply-back rates. We found that no general users engaged with the rumour post before and after the guardian intervention. The most significant differences in the emotive characteristics of replies were across topics, with rumour starters expressing more anger, disgust, and offensive language when discussing rumours about terrorism and racism. These emotions indicate the backfire effect. Our findings raise concerns about the effectiveness of fact-checking interventions in updating misinformed users. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Tweets

来源：评论

学校读者我要写书评

暂无评论

Blockchain-enhanced artificial intelligence for advanced collision avoidance in the Internet of Vehicles (IoV)

引用

Neural Computing and Applications 2025年第6期37卷 4915-4936页

作者： Talaat, Fatma M. Hamza, Alyaa A. Faculty of Artificial Intelligence Kafrelsheikh University Kafrelsheikh33516 Egypt Faculty of Engineering Technology Elsewedy University of Technology–Polytechnic of Egypt 10th of Ramadan City44629 Egypt Faculty of Computer Science & Engineering New Mansoura University Gamasa35712 Egypt Bahrain Polytechnic Isa Town Bahrain Badr City11829 Egypt

The convergence of blockchain technology and artificial intelligence (AI) presents a promising solution for enhancing safety within the Internet of Vehicles (IoV) ecosystem. This paper introduces the "Blockchain-Based Collision Avoidance with AI for Vehicles" (BCA-CAR) algorithm, which aims to provide advanced and intelligent collision avoidance capabilities in IoV. BCA-CAR combines the security and data integrity features of blockchain with the real-time decision-making capabilities of AI to prevent collisions and improve road safety. The algorithm consists of five key phases: Data Collection and Processing, AI Collision Risk Assessment, Decision and Smart Contract Execution, Data Validation and Trust (Blockchain Integration), and Learning and Improvement. In the Data Collection and Processing phase, data from vehicle sensors, cameras, V2V and V2I communication, and external infrastructure is collected and preprocessed. The AI Collision Risk Assessment phase utilizes machine learning models to analyze real-time data and predict collision risks. In the Decision and Smart Contract Execution phase, smart contracts on the blockchain automate collision avoidance actions. The Data Validation and Trust phase ensures the authenticity and integrity of data through blockchain technology. Finally, the Learning and Improvement phase leverages historical collision data to enhance predictive models and overall system performance. BCA-CAR's primary objective is to enhance safety by preventing collisions, ensuring data trustworthiness, and providing intelligent collision avoidance capabilities. This innovative algorithm has the potential to revolutionize road safety in the era of IoV by reducing accidents, improving traffic management, and enhancing the security and privacy of vehicular communication. The findings highlight that Support Vector Regression (SVR) demonstrates strong predictive accuracy and adaptability within the Internet of Vehicles (IoV), offering a reliable modeli

关键词： Risk assessment

来源：评论

学校读者我要写书评

暂无评论

Portable Blockchain for System Testing of Solana Smart Contracts 24

Portable Blockchain for System Testing of Solana Smart Contr...

引用

7th International Conference on Blockchain technology and Applications, ICBTA 2024

作者： Madhwal, Yash Yanovich, Yury Zhintiy, Grigoriy Project Center for Next Generation Wireless and IoT Skolkovo Institute of Science and Technology Moscow Russia Faculty of Computer Science Hse University Moscow Russia

ISBN: (纸本)9798400717390

This paper presents an in-depth analysis of the architecture and development of an innovative application designed to virtualize transaction execution on the Solana blockchain. This virtualization system facilitates the execution of smart contracts at native speeds, while concurrently evaluating the outcomes within a controlled and isolated environment, separate from the live blockchain network. By employing this approach, the system maximizes performance efficiency and ensures the integrity of blockchain data during testing phases. The study meticulously examines the key design considerations, development challenges, and technological implementations in constructing this sophisticated virtualization tool. Furthermore, the application forms the foundation for a comprehensive testing suite that rigorously assesses the performance and security of smart contracts before their deployment on the Solana network. This work aims to advance the blockchain development field by offering a robust framework for testing and validation, thereby enhancing the reliability of smart contracts through rigorous pre-deployment evaluation processes. © 2024 Copyright held by the owner/author(s).

关键词： Blockchain

来源：评论

学校读者我要写书评

暂无评论

Gas Consumption Minimization for Dividend Payout Smart Contract 24

Gas Consumption Minimization for Dividend Payout Smart Contr...

引用

7th International Conference on Blockchain technology and Applications, ICBTA 2024

作者： Yanovich, Yury Shuklin, Maksim Madhwal, Yash Computer Science Faculty Hse University Moscow Russia Project Center for Next Generation Wireless and IoT Skolkovo Institute of Science and Technology Moscow Russia

ISBN: (纸本)9798400717390

Investors acquire securities to gain a share of the profits generated by the underlying assets. These profits are distributed among the investors in the form of dividends. With the advent of decentralized networks, executing smart contracts has become an effective way to implement dividend payments. This work aims to develop algorithms that facilitate efficient dividend payments while minimizing the gas consumed in various automated scenarios. The algorithms have been implemented in the Solidity programming language specifically designed for smart contracts. This study proposes several algorithms, each with unique advantages, which have been demonstrated for different scenarios. The main objective of this work is to provide investors with a reliable and efficient method for receiving their dividends in a decentralized manner. © 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.

关键词： Ethereum

来源：评论

学校读者我要写书评

暂无评论

Video classification for efficient data storage using deep learning: a comparison of sequential and simultaneous feature extraction methods

引用

Multimedia Tools and Applications 2025年第6期84卷 3071-3094页

作者： Dang Thi, Phuc Nguyen Dang, Hoang Thi Doan Huu, Phuc Dau Sy, Hieu Department of Computer Science Faculty of Information Technology Industrial University of Ho Chi Minh City 12 Nguyen Van Bao Ho Chi Minh City7000 Viet Nam Department of Applied Physics Faculty of Applied Science University of Technology - Viet Nam National University HCMC To Hien Thanh Ho Chi Minh City7000 Viet Nam

Optimizing camera information storage is a critical issue due to the increasing data volume and a large number of daily surveillance videos. In this study, we propose a deep learning-based system for efficient data storage. Videos captured by cameras are classified into four categories: no action, normal action, human action, and dangerous action. Videos without action or with normal action are stored temporarily and then deleted to save storage space. Videos with human action are stored for easy retrieval, while videos with dangerous action are promptly alerted to users. In the paper, we propose two approaches using deep learning models to address the video classification problem. The first approach is a separate approach, where pretrained CNN models extract features from video frame images. These features are then passed through RNN, Transformer models to extract relationships between them. The goal of this approach is to delve into extracting features of objects in the video. The proposed models include VGG16, InceptionV3 combined with LSTM, BiLSTM, Attention, and Vision Transformer. The next approach combines CNN and LSTM layers simultaneously through models like ConvLSTM and LRCN. This approach aims to help the model simultaneously extract object features and their relationships, with the goal of reducing model size, accelerating the training process, and increasing object recognition speed when deployed in the system. In Approach 1, we construct and refine network architectures such as VGG16+LSTM, VGG16+Attention+LSTM, VGG16+BiLSTM, VGG16+ViT, InceptionV3+LSTM, InceptionV3+Attention+LSTM, InceptionV3+BiLSTM. In Approach 2, we build a new network architecture based on the ConvLSTM and LRCN model. The training dataset, collected from real surveillance cameras, comprises 3315 videos labeled into four classes: no action (1018 videos), actions involving people (832 videos), dangerous actions (751 videos), and normal actions (714 videos). Experimental results show t

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Behaviour-diverse automatic penetration testing:a coverage-based deep reinforcement learning approach

引用

Frontiers of computer science 2025年第3期19卷 15-24页

作者： Yizhou YANG Longde CHEN Sha LIU Lanning WANG Haohuan FU Xin LIU Zuoning CHEN Zhongguancun Laboratory Beijing 100081China Zhejiang Lab Hangzhou 311121China National Research Centre of Parallel Computer Engineering and Technology Wuxi 214000China Faculty of Geographical Science Beijing Normal UniversityBeijing 100875China Department of Earth System Science Tsinghua UniversityBeijing 100084China

Reinforcement Learning(RL)is gaining importance in automating penetration testing as it reduces human effort and increases ***,given the rapidly expanding scale of modern network infrastructure,the limited testing scale and monotonous strategies of existing RLbased automated penetration testing methods make them less effective in practical *** this paper,we present CLAP(Coverage-Based Reinforcement Learning to Automate Penetration Testing),an RL penetration testing agent that provides comprehensive network security assessments with diverse adversary testing behaviours on a massive *** employs a novel neural network,namely the coverage mechanism,to address the enormous and growing action spaces in large *** also utilizes a Chebyshev decomposition critic to identify various adversary strategies and strike a balance between *** results across various scenarios demonstrate that CLAP outperforms state-of-the-art methods,by further reducing attack operations by nearly 35%.CLAP also provides enhanced training efficiency and stability and can effectively perform pen-testing over large-scale networks with up to 500 ***,the proposed agent is also able to discover pareto-dominant strategies that are both diverse and effective in achieving multiple objectives.

关键词： network security penetration testing reinforcement learning artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

YOLOv8n-CGW: A novel approach to multi-oriented vehicle detection in intelligent transportation systems

引用

Multimedia Tools and Applications 2025年第7期84卷 3809-3840页

作者： Berwo, Michael Abebe Fang, Yong Sarwar, Nadeem Mahmood, Jabar Aljohani, Mansourah Elhosseini, Mostafa School of Information and Engineering Chang’an University Shaanxi Xi’an710064 China Department of Computer Science Bahria University Lahore Campus Punjab Lahore54600 Pakistan Department of Information Technology Faculty of Computing and Information Technology University of Sialkot Punjab Sialkot51040 Pakistan College of Computer Science and Engineering Taibah University Yanbu46421 Saudi Arabia Computers and Control Systems Engineering Department Faculty of Engineering Mansoura University Mansoura35516 Egypt

In the context of Intelligent Transportation Systems (ITS), the role of vehicle detection and classification is indispensable for streamlining transportation management, refining traffic control, and conducting in-depth accident analyses. However, the intricate task of accurately detecting multi-oriented vehicles in diverse scenarios remains a challenge, even with the advancements in ITS. Factors such as vehicle morphology, design, road conditions, structural details, and climatic variables provide a significant part in the detection. To address this, our study introduces an advanced YOLOv8n model, specifically the YOLOv8n-CGW, which is built upon pre-trained deep learning frameworks. This model is meticulously tailored to enhance the detection capabilities for multi-oriented vehicles. Comprehensive evaluations on esteemed datasets like MtV, MERGED, Custom dataset, DAWN, WEDGE, and UA-DETRAC underscore the superiority of our approach over existing models. Notably, our YOLOv8n-CGW model achieves a remarkable mAP of 83.3% and boasts a swift inference time of 0.0019s, setting a new benchmark in the domain. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Deep neural networks

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：