检索结果-内蒙古大学图书馆

Just: towards jute pest classification by combination of supervised learning and triplet loss aided contrastive learning

引用

Iran Journal of computer science 2025年 1-14页

作者： Kundu, Shreyan Mukhopadhyay, Souradeep Talukdar, Rahul Das, Semanti Adhikari, Subhajit Computer Science and Engineering (Artificial Intelligence) Institute of Engineering and Management Iem Management Building D-1 Street No. 13 EP Block Sector V Bidhannagar West Bengal Kolkata 700091 India Computer Science and Automation Indian Institute of Science CV Raman Road Karnataka Bangalore 560012 India Computer Science and Engineering (core) Institute of Engineering and Management Iem Management Building D-1 Street No. 13 EP Block Sector V Bidhannagar West Bengal Kolkata 700091 India Department of Lifescience Christ University Hosur Main Road Bhavani Nagar Karnataka Bangalore 560029 India Department of Information Technology RCCIIT KOLKATA West Bengal Kolkata 700015 India

Jute is a vital agricultural commodity contributing significantly to the GDP of countries like Bangladesh, India, Myanmar, and China. However, because of its inaccuracy and slowness, its vulnerability to pest infestations-which are often handled by manual inspections-poses serious cost concerns. This study suggests a unique method for early and accurate pest identification that combines contrastive and supervised learning. Contrastive learning enhances feature representation by distinguishing between positive and negative samples, ensuring that instances within the same class are closely grouped while maintaining separation between different classes. It reduces false negatives by classifying some samples as negative and others with the same label as positive. Supervised learning enables precise pest identification by aligning features with distinctive characteristics of each class. Metrics including precision, recall, F1 score, ROC curve, and confusion matrix are used to assess the hybrid model’s performance;the findings show notable accuracy gains over conventional techniques. This scalable and dependable solution lowers losses caused by pests and provides a sustainable method of growing jute using cutting-edge advanced machine-learning techniques. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.

关键词： Augmentation Contrastive learning Supervised learning Triplet loss

来源：评论

学校读者我要写书评

暂无评论

Empowering Adaptive Endogenous Security Trend Prediction Detection for IoT Sensor Nodes 5

Empowering Adaptive Endogenous Security Trend Prediction Det...

引用

5th International Conference on Advances in Electrical, Electronics and Computing Technology, EECT 2025

作者： Zhou, Lihu Dong, Xiuying Zhang, Enli Wang, Ting Zhang, Xiao Zhang, Chong School of Computer Science and Software Engineering Southwest Petroleum University Chengdu China Exploration and Development Research Institute of PetroChina Southwest Oil & Gasfield Company Chengdu China School of Computer Science Chengdu University of Information Technology Chengdu China

ISBN: (纸本)9798331541545

Massively deployed IoT devices require a lightweight design to reduce costs. However, this architecture inherently limits their security, increasing the risk of data breaches and tampering on a large scale. In this paper, we propose CGAD, a model that leverages one-dimensional convolution (Conv1D), gated recurrent units (GRU), and the attention mechanism to analyze multi-sensor time series data at the data gateway from sensing terminals. To achieve this, we generate time series predictions and detect anomalies by comparing predicted values with real data using a threshold-based approach, thereby enhancing data security. To address the labeling problem of massive data, we employ unsupervised learning and design a dynamic sliding window threshold to improve the judgment of time series with varying characteristics. We also used a number of datasets, such as the KW51 railroad bridge dataset and the self-picked temperature and humidity dataset, to evaluate our system. The results demonstrate that our CGAD model achieves recall rates of 98% and 99% for anomaly detection on the two datasets, respectively. Additionally, the precision and F1 scores surpass other methods, ensuring cost-effectiveness for massive terminals and addressing the security issues of sensitive data. These advantages contribute to realizing the Internet of Everything. © 2025 IEEE.

关键词： Gateways (computer networks)

来源：评论

学校读者我要写书评

暂无评论

Deep-EERA:DRL-Based Energy-Efficient Resource Allocation in UAV-Empowered Beyond 5G Networks

引用

Tsinghua science and Technology 2025年第1期30卷 418-432页

作者： Shabeer Ahmad Jinling Zhang Ali Nauman Adil Khan Khizar Abbas Babar Hayat School of Electronic Engineering Beijing University of Posts and TelecommunicationBeijing 1000876China Department of Information and Communication Engineering Yeungnam UniversityGyeongsan 38541Republic of Korea School of Information Engineering Xi’an Eurasia UniversityXi’an 710065China Department of Computer Science Hanyang UniversitySeoul 04763Republic of Korea

The rise of innovative applications,like online gaming,smart healthcare,and Internet of Things(IoT)services,has increased demand for high data rates and seamless connectivity,posing challenges for Beyond 5G(B5G)*** is a need for cost-effective solutions to enhance spectral efficiency in densely populated areas,ensuring higher data rates and uninterrupted connectivity while minimizing *** Aerial Vehicles(UAVs)as Aerial Base Stations(ABSs)offer a promising and cost-effective solution to boost network capacity,especially during emergencies and high-data-rate ***,integrating UAVs into the B5G networks presents new challenges,including resource scarcity,energy efficiency,resource allocation,optimal power transmission control,and maximizing overall *** paper presents a UAV-assisted B5G communication system where UAVs act as ABSs,and introduces the Deep Reinforcement Learning(DRL)based Energy Efficient Resource Allocation(Deep-EERA)*** efficient DRL-based Deep Deterministic Policy Gradient(DDPG)mechanism is introduced for optimal resource allocation with the twin goals of energy efficiency and average throughput *** proposed Deep-EERA method learns optimal policies to conserve energy and enhance throughput within the dynamic and complex UAV-empowered B5G *** extensive simulations,we validate the performance of the proposed approach,demonstrating that it outperforms other baseline methods in energy efficiency and throughput maximization.

关键词： Deep Reinforcement Learning(DRL) Unmanned Aerial Vehicles(UAVs) resource allocation energy efficiency 5G and beyond network

来源：评论

学校读者我要写书评

暂无评论

Online payment fraud monitoring and detection: Performance analysis of tree-based Ensemble Machine Learning models

Online payment fraud monitoring and detection: Performance a...

引用

International Communication Systems and Networks and Workshops, COMSNETS

作者： Somnath Bhowmik Jaydeep Howlader Computer Science and Engineering NIT Durgapur India

ISBN: (数字)9798331531195

ISBN: (纸本)9798331531201

In today’s world of digital transformation in business, driven by rapid technological advancement, the use of electronic commerce and digital banking is ever increasing. In line with global trends, the Indian banking and financial sector is experiencing a major adoption of digital payments, be it smaller retail payments or high-value corporate payments. With the increasing use of online payments, money laundering, cyberattacks, and other fraudulent activities are increasing. The innovative use of technology by fraudsters, aided by dynamic and irregular patterns of fraudulent activities, poses additional challenges in identifying frauds with high accuracy. Hence, preventing financial fraud requires the implementation of secured payment systems and efficient fraud detection and monitoring *** study delves into analyzing the results of applying ensemble-based Machine Learning (ML) methods to digital payment transactions, aiming to enhance the accuracy of fraud detection. A recent study and review of the existing literature show better performance by the Decision Tree algorithm over other ML models. Taking this into account, we used the decision tree algorithm to further build tree-based ensemble models, namely Random Forest, Gradient Tree Boosting, AdaBoost, and XGBoost. We analyze seven evaluation metrics: Precision, Recall, F-score, Accuracy, Misclassification Error, Area Under the Receiver Operating Characteristic Curve, and Cohen Kappa score. The data sets used are the National Electronic Fund Transfer (NEFT) payment transactions. The results show encouraging scores in segregating valid and fraudulent transactions, and thereby suggest the use of tree-based ensemble models as an effective tool for continuous monitoring and early detection of fraudulent payment transactions.

关键词： Analytical models Accuracy Machine learning algorithms Banking Predictive models Data models Fraud Monitoring Random forests Classification tree analysis

来源：评论

学校读者我要写书评

暂无评论

Federated Class-Incremental Learning via Weighted Aggregation and Distillation

引用

IEEE Internet of Things Journal 2025年

作者： Wu, Feng Tan, Alysa Ziying Feng, Siwei Yu, Han Deng, Tao Zhao, Libang Chen, Yuanlu Soochow University School of Computer Science and Technology Suzhou215000 China Nanyang Technological University School of Computer Science and Engineering 639798 Singapore

Federated Class-Incremental Learning (FCIL) aims to design privacy-preserving collaborative training methods to continuously learn new classes from distributed datasets. In these scenarios, federated clients face the challenge of encountering new classes while being constrained by limited memory capacity, which can lead to catastrophic forgetting in the resulting global model. Existing FCIL approaches tend to overlook the challenges posed by the heterogeneity of dataset label distribution among clients, thereby constraining the generalization capacity of the global model they learn. Some of these methods also suffer from excessive computational burdens when addressing catastrophic forgetting problems. Furthermore, certain approaches are constrained to handling only straightforward data, posing significant difficulties in managing complex datasets and tackling more intricate scenarios. In this paper, we propose the Weighted Aggregation and Distillation-based FCIL (WAD-FCIL) method to address these limitations. To address data heterogeneity arising from class imbalance, we first introduce a task-aware client clustering method to identify clients with extreme class deviations before global model aggregation to eliminate potential impact on the global model. Then, we propose a multi-sampling weighted aggregation approach during the global FL model update that integrates knowledge from different clients and dynamically adjusts the weight of each client model to facilitate model update. To mitigate catastrophic forgetting, we propose a multi-model distillation strategy that involves selecting multiple teacher models for knowledge distillation. Extensive experiments comparing WAD-FCIL with ten state-of-the-art methods demonstrate that it significantly outperforms the baselines by 0.8%-3.2% in terms of average test accuracy on three representative benchmark datasets. © 2025 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Proposing a Secure Framework for Industrial AI Security in 21st Century

Proposing a Secure Framework for Industrial AI Security in 2...

引用

Robotics and Technologies for Industrial Automation (ROBOTHIA), IEEE International Conference on

作者： Venkata Venugopal Rao Gudlur University Malaysia of Computer Science & Engineering (UNIMY)

ISBN: (数字)9798350356755

ISBN: (纸本)9798350356762

This proposed solution seeks to establish a secure framework for implementing Industrial Artificial Intelligence (AI) Security in the 21st century. As AI technology progresses through creative and innovative advancements, ensuring robust security and addressing system vulnerabilities is essential for maintaining trust among online users. The proposed framework will understand and prevent key security concerns, such as data privacy, algorithmic transparency, and resilience against cyber threats, to create a more secure environment for the future. AI is deeply integrated into sectors like finance, healthcare, and autonomous systems related to various industries in the 21st century. This framework seeks to balance innovation with security, fostering trust and reliability in AI-powered solutions for industries. Significant Observations are Growing Threat Landscape, Trust Gap in AI Adoption, Lack of Standardized Security Protocols. The Quantifying Results are Attack Resistance, Data Privacy Assurance, System Reliability Improvement, Enhanced User Trust and Adoption. Securing web applications is particularly challenging due to the difficulty in detecting and recognizing user behavior patterns that could signal a potential compromise or cyber threat. By adopting robust security protocols, this framework will facilitate the safe and ethical integration of AI into various online applications, fostering trust and driving innovation in AI-driven solutions.

关键词： Training Industries Technological innovation Ethics Protocols Service robots Security Reliability Artificial intelligence Monitoring

来源：评论

学校读者我要写书评

暂无评论

Interpretable Arabic Hate Speech Detection using Large Language Model

Interpretable Arabic Hate Speech Detection using Large Langu...

引用

Advanced Innovations in Smart Cities (ICAISC), International Conference on

作者： Majed Alwateer Ibrahim Gad Mahmoud Elmarhomy Ghada Elmarhomy Hasan Hashim Malik Almaliki El-Sayed Atlam Department of Computer Science College of Computer Science and Engineering Taibah University Saudi Arabia Computer Science Department Faculty of Science Tanta University Tanta Egypt Computer Science and Information College of Applied Taibah University Saudi Arabia Department of Information Systems College of Computer Science and Engineering Taibah University Saudi Arabia

ISBN: (数字)9798331506995

ISBN: (纸本)9798331507008

Social media platforms serve as significant spaces for users to have conversations, discussions and express their opinions. However, anonymity provided to users on these platforms allows the spread of hate speech and other offensive material. Due to the wide-ranging nature of these platforms, there is a critical need to automatically detect and report occurrences of hate speech. There are various detection methods, but many of them operate as black boxes, lacking interpretability and explainability by design. To address the lack of interpretability, this study explores the development of an interpretable framework to detect hate speech in Arabic using large language models (LLMs). The proposed approach combines advanced natural language processing techniques with interpretable machine learning methods to enhance understanding of model decisions. The experimental results demonstrate that the model achieves high accuracy while maintaining interpretability, enabling users to understand the reasoning behind the detections. The proposed method achieves an accuracy of 0.846%, with a precision of 0.843% and a recall of 0.846%, outperforming existing Arabic hate speech detection models. These results show the effectiveness of combining LLM with interpretability for this critical task, providing a reliable and transparent solution for automated moderation of harmful content.

关键词： Technological innovation Accuracy Social networking (online) Smart cities Large language models Hate speech Text categorization Writing Reliability Stakeholders

来源：评论

学校读者我要写书评

暂无评论

Co-Designing Binarized Transformer and Hardware Accelerator for Efficient End-to-End Edge Deployment 24

Co-Designing Binarized Transformer and Hardware Accelerator ...

引用

43rd International Conference on computer-Aided Design, ICCAD 2024

作者： Ji, Yuhao Fang, Chao Ma, Shaobo Shao, Haikuo Wang, Zhongfeng School of Electronic Science and Engineering Nanjing University Nanjing China Department of Computer Science and Engineering Chinese University of Hong Kong Hong Kong School of Integrated Circuits Sun Yat-sen University Shenzhen China

ISBN: (纸本)9798400710773

Transformer models have revolutionized AI tasks, but their large size hinders real-world deployment on resource-constrained and latency-critical edge devices. While binarized Transformers offer a promising solution by significantly reducing model size, existing approaches suffer from algorithm-hardware mismatches with limited co-design exploration, leading to suboptimal performance on edge devices. Hence, we propose a co-design method for efficient end-to-end edge deployment of Transformers from three aspects: algorithm, hardware, and joint optimization. First, we propose BMT, a novel hardware-friendly binarized Transformer with optimized quantization methods and components, and we further enhance its model accuracy by leveraging the weighted ternary weight splitting training technique. Second, we develop a streaming processor mixed binarized Transformer accelerator, namely BAT, which is equipped with specialized units and scheduling pipelines for efficient inference of binarized Transformers. Finally, we co-optimize the algorithm and hardware through a design space exploration approach to achieve a global trade-off between accuracy, latency, and robustness for real-world deployments. Experimental results show our co-design achieves up to 2.14~49.37× throughput gains and 3.72~88.53× better energy efficiency over state-of-the-art Transformer accelerators, enabling efficient end-to-end edge deployment. © 2024 Copyright is held by the owner/author(s).

关键词： Integrated circuit layout

来源：评论

学校读者我要写书评

暂无评论

Crime Pulse: Crime hot-spot detection using Machine Learning and Gen-AI*

Crime Pulse: Crime hot-spot detection using Machine Learning...

引用

Computational Intelligence, Communication Technology and Networking (CICTN), International Conference on

作者： Srijan Singh Vandit Singh Aditi Arora Department of Computer Science and Engineering ABESEC Ghaziabad

ISBN: (数字)9798331530389

ISBN: (纸本)9798331530396

The aim of this research is to determine the accuracy of machine learning algorithms in identifying high-crime regions. Traditional methods of locating crime hotspots, such manual analysis, are challenging and prone to errors. Machine learning offers a more accurate and effective approach. The study procedure includes preprocessing, determining pertinent characteristics, and collecting crime data. A range of machine learning algorithms that predict the likelihood of crime in different places will be trained using this data. The performance of these models will be evaluated using the pertinent metrics. The expected outcomes include demonstrating the value of machine learning in identifying crime hotspots, identifying suitable algorithms, gaining a better understanding of the factors that lead to hotspots, and developing a practical tool for law enforcement. The goal of this research is to make society safer and more secure by utilizing machine learning. Law enforcement organizations may more effectively deploy resources, discourage criminal activity, and eventually lower crime rates by accurately and efficiently identifying high-crime areas.

关键词： Ethics Machine learning algorithms Accuracy Law enforcement Prevention and mitigation Machine learning Manuals Prediction algorithms Public security Resource management

来源：评论

学校读者我要写书评

暂无评论

Rebalanced Multimodal Learning with Data-aware Unimodal Sampling

arXiv

引用

arXiv 2025年

作者： Jiang, Qingyuan Chi, Zhouyang Ma, Xiao Mao, Qirong Yang, Yang Tang, Jinhui School of Computer Science and Engineering Nanjing University of Science and Technology China School of Computer Science and Technology Zhejiang Sci-Tech University China School of Computer Science and Communication Engineering Jiangsu University China

To address the modality learning degeneration caused by modality imbalance, existing multimodal learning (MML) approaches primarily attempt to balance the optimization process of each modality from the perspective of model learning. However, almost all existing methods ignore the modality imbalance caused by unimodal data sampling, i.e., equal unimodal data sampling often results in discrepancies in informational content, leading to modality imbalance. Therefore, in this paper, we propose a novel MML approach called Data-aware Unimodal Sampling (DUS), which aims to dynamically alleviate the modality imbalance caused by sampling. Specifically, we first propose a novel cumulative modality discrepancy to monitor the multimodal learning process. Based on the learning status, we propose a heuristic and a reinforcement learning (RL)based data-aware unimodal sampling approaches to adaptively determine the quantity of sampled data at each iteration, thus alleviating the modality imbalance from the perspective of sampling. Meanwhile, our method can be seamlessly incorporated into almost all existing multimodal learning approaches as a plugin. Experiments demonstrate that DUS can achieve the best performance by comparing with diverse state-of-the-art (SOTA) baselines. © 2025, CC BY.

关键词： Contrastive 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：