检索结果-内蒙古大学图书馆

International Conference on Data science and Advanced Analytics (DSAA)

作者： Damien A. Dablain Nitesh V. Chawla Dept. Computer Science and Engineering University of Notre Dame Notre Dame USA

ISBN: (数字)9798350364941

ISBN: (纸本)9798350364958

Real-world data is often imbalanced, such that the number of training instances varies by class. Data augmentation (DA) of under-represented classes is commonly used to improve model generalization in the face of class imbalance. Despite its ubiquity, the impact of data augmentation on machine learning (ML) models is not clearly understood. Here, we undertake a holistic examination of the effect of DA on under-represented classes. Unlike other studies, which focus on a single ML model type, we examine three different classifier families: convolutional neural networks, support vector machines, and logistic regression models; five different DA techniques and two different data modalities - image and tabular. Our research indicates that DA, when applied to imbalanced data, produces substantial changes in model weights, support vectors and front-end feature selection. These changes occur with respect to all classes, not just the ones that DA is applied to. Further, our empirical analysis shows that data augmentation's positive influence on generalization does not necessarily occur as a result of reducing weight norms. Rather, weight and support vector specialization play important roles in generalization. The specialization process may be a form of memorization that is spawned by variances introduced by augmented data. We investigate the seeming contradiction between improved generalization versus weight and support vector specialization.

关键词： Training Support vector machines Logistic regression Neural networks Machine learning Data augmentation Feature extraction Data models Vectors Faces

来源：评论

学校读者我要写书评

暂无评论

An EEG-Based User Authentication System Using Event-Related Potentials and Ensemble Learning

An EEG-Based User Authentication System Using Event-Related ...

引用

Cyber Awareness and Research Symposium (CARS)

作者： Soudabeh Bolouri Diksha Shukla Dept. of Electrical Engineering & Computer Science University of Wyoming Laramie WY USA

ISBN: (数字)9798350386417

ISBN: (纸本)9798350386424

This paper presents an EEG-based user authentication system using Event-Related Potentials (ERPs) to distinguish legitimate users from impostors. Utilizing a publicly available EEG dataset, we implemented a comprehensive data processing pipeline, which included advanced preprocessing and feature extraction techniques. Multiple state-of-the-art machine learning classifiers, such as CatBoost and XGBoost, were evaluated to assess their effectiveness in user authentication. The results showed a very low average Equal Error Rate (EER) of 2.53%. Our study emphasizes the strength of the P300 and N400 responses in biometric authentication and demonstrates the potential of advanced ensemble classifiers in improving system accuracy. This research contributes to the development of EEG-based authentication and lays the groundwork for future studies aiming to create secure and practical biometric systems.

关键词： Accuracy Authentication Signal processing Feature extraction Biometric authentication Robustness Electroencephalography Real-time systems Security Time-domain analysis

来源：评论

学校读者我要写书评

暂无评论

DeepLeaf-Attentive: Integrating Attention Mechanisms for Crop Leaf Disease Identification

DeepLeaf-Attentive: Integrating Attention Mechanisms for Cro...

引用

International Conference on computer and Information Technology (ICCIT)

作者： Tanvir Ahmed Farzana Sharmin Mou Dept. of Computer Science Engineering Dhaka University of Engineering & Technology Gazipur Dhaka Bangladesh Dept. of Computer Science University of Birmingham Birmingham UK

Crop leaf diseases have a detrimental impact on agricultural production. Visual inspection by organic farmers for disease detection is time-consuming and ineffective. To address this, we propose DeepLeaf-Attentive, a novel deep learning model for classifying crop leaf diseases. It combines MobileNet architecture with additional layers and attention mechanism. It is trained on a dataset of 30,000 images including data augmentation representing 60 classes with 500 images per class. The model achieves outstanding performance, with 97.21% accuracy, 97.72% precision, 97.43% recall, 97.49% F1-score and 18.0 ms inference time. Comparative evaluations against state-of-the-art approaches demonstrate the superiority of DeepLeaf-Attentive in crop leaf disease classification. By providing a reliable and efficient tool, our model will empower farmers to accurately identify and categorize various crop leaf diseases, facilitating informed decision-making and enabling the implementation of targeted and timely treatment strategies. The adoption of DeepLeaf-Attentive can significantly improve disease management practices in agriculture, leading to enhanced crop yields, reduced economic losses, and sustainable farming practices. By revolutionizing crop leaf disease identification, our research contributes to the advancement of precision agriculture, benefiting farmers and supporting global food security.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Earthquake Early Magnitude Prediction Using 3D Point Cloud Based Deep Learning Model: Case of 2018 Osaka Earthquake

Earthquake Early Magnitude Prediction Using 3D Point Cloud B...

引用

Trends in Electrical, Electronics, computer engineering Conference (TEECCON)

作者： Anushka Joshi Balasubramanian Raman Dept. of Computer Science and Engineering Indian Institute of Technology Roorkee Roorkee India

ISBN: (数字)9798331505714

ISBN: (纸本)9798331505721

Earthquakes are considered to be one of the deadliest natural phenomena that exist on Earth. The destruction during an earthquake is directly related to the magnitude corresponding to a station during the earthquake. The magnitude of an earthquake is calculated from the complete time series recorded by the instrument. In recent years, the early phase of the time series has been used to measure an earthquake’s magnitude, commonly known as an earthquake early warning system. This early detection of magnitude gives sufficient time for saving lives and property damaged by the earthquake. Traditionally, simple linear relations are utilized in the EEWS. In this study, the proposed Point Cloud Magnitude Prediction (CloudMag) model utilized point cloud technology, which converts two-dimensional time series data into multi-dimensional representation in space. This representation is then utilized for earthquake magnitude prediction using a convolutional neural network architecture. This study discusses a case of the deadly Osaka earthquake that occurred on 18 June 2018 at 7:58:35 AM Japan Standard Time. The result shows that the novel deep learning approach for magnitude estimation is superior to traditional and other machine learning models.

关键词： Point cloud compression Deep learning Solid modeling Three-dimensional displays Computational modeling Earthquakes Time series analysis Predictive models Time measurement Standards

来源：评论

学校读者我要写书评

暂无评论

Water Region Extraction in Aerial Images Using Deep Learning

Water Region Extraction in Aerial Images Using Deep Learning

引用

Intelligent Computing and Sustainable Innovations in Technology (IC-SIT), International Conference on

作者： Nikita Pandey Saumya Kumar Satish Kumar Yadav Shashwat Srivastava Shomil Singh Deepak Gupta Dept. of Computer Science and Engineering Pranveer Singh Institute of Technology Kanpur India

ISBN: (数字)9798350369175

ISBN: (纸本)9798350369182

Water bodies are essential component of Earth’s surface and accurate water region mapping is crucial for disaster management and environmental monitoring. Existing methods to map water bodies relies on traditional modelling methods that need extensive manual calibration struggle with scalability and lack advanced pattern recognition. This paper evaluates the effectiveness of deep learning-based semantic segmentation models, DeepLabV3+ and U-Net, using aerial images resized to $\mathbf{2 5 6 x} 256$ pixels in mapping water regions in aerial images. DeepLabV3+, with a ResNet50 backbone, and U-Net are trained on a comprehensive flood-affected dataset with preprocessing techniques such as resizing and augmentation. Model performance is assessed using metrics like IoU, F1 score, accuracy, precision, and recall. The results indicate that DeepLabV3+ outperforms U-Net, showing promise for automating water region mapping and improving flood surveys and urban planning.

关键词： Training Measurement Deep learning Surveys Technological innovation Accuracy Computational modeling Semantic segmentation Urban planning Water resources

来源：评论

学校读者我要写书评

暂无评论

Adaptive Traffic Signal Management Using Real-Time Vehicle Detection and Tracking

Adaptive Traffic Signal Management Using Real-Time Vehicle D...

引用

Interdisciplinary Approaches in Technology and Management for Social Innovation (IATMSI), IEEE Conference on

作者： Sarvesh Shirulkar Rishabhdev Makode Richa Khandelwal Dept. of Electronics and Computer Science Shri Ramdeobaba College of Engineering and Management Nagpur India Dept. of Electronics Engineering Ramdeobaba University Nagpur India

ISBN: (数字)9798331521691

ISBN: (纸本)9798331521707

Traffic congestion is a significant issue in urban cities, as it restricts traffic flow, increases fuel consumption, and produces high emissions. Traditional traffic signal systems work on fixed timing without considering real-time variations in traffic and often worsen the problem. This paper describes an adaptive traffic signal control system based on the YOLO (You Only Look Once) algorithm, using object detection for real-time signal timing in response to real-time traffic. The system applies real-time vehicle detection and tracking at intersections for dynamic management of green light time according to volume and vehicle type. The proposed approach makes vehicle recognition significantly efficient and responds quickly to the change in traffic situation. The paper discusses the design, implementation, and evaluation of performance in a simulated environment. The results demonstrate improved traffic management and reduced congestion, offering a promising solution to smart city traffic control.

关键词： YOLO Adaptive systems Smart cities Vehicle detection Green products Traffic control Real-time systems Timing Dynamical systems Vehicle dynamics

来源：评论

学校读者我要写书评

暂无评论

Plant Disease Detection using Convolutional Neural Network

Plant Disease Detection using Convolutional Neural Network

引用

Smart Devices (ICSD), International Conference on

作者： Jyoti Agarwal Shivansh Naithani Utkarsh Singh Harsh Raj Dept. of Computer Science and Engineering Graphic Era Deemed to be University Dehradun India

ISBN: (数字)9798350394412

ISBN: (纸本)9798350394429

Pest-infested plants and crops have a negative effect on the nation's agricultural output. To detect and identify diseases in plants, farmers or other experts must keep a careful watch on plants. However, this method is frequently expensive, time-consuming, and inaccurate. Finding a spot on the leaves of the infected plant is one method of detecting plant diseases. This research aims to develop a disease detection model based on the classification of leaf images. To detect plant illnesses the proposed model utilizes image processing with a convolution neural network (CNN). This research work focus on three different plant diseases: tomato, corn and potato. Experimental results show high accuracy rate of 98.8%. Proposed work is also compared with some existing work working o same plants diseases.

关键词： Plant diseases Convolution Plants (biology) Image processing Neural networks Crops Forestry Vectors Convolutional neural networks Smart devices

来源：评论

学校读者我要写书评

暂无评论

Remote Training in Task-Oriented Communication: Supervised or Self-Supervised with Fine-Tuning?

arXiv

引用

arXiv 2025年

作者： Li, Hongru Zhao, Hang He, Hengtao Song, Shenghui Zhang, Jun Letaief, Khaled B. Dept. of Electronic and Computer Engineering The Hong Kong University of Science and Technology Hong Kong

Task-oriented communication focuses on extracting and transmitting only the information relevant to specific tasks, effectively minimizing communication overhead. Most existing methods prioritize reducing this overhead during inference, often assuming feasible local training or minimal training communication resources. However, in real-world wireless systems with dynamic connection topologies, training models locally for each new connection is impractical, and task-specific information is often unavailable before establishing connections. Therefore, minimizing training overhead and enabling label-free, task-agnostic pre-training before the connection establishment are essential for effective task-oriented communication. In this paper, we tackle these challenges by employing a mutual information maximization approach grounded in self-supervised learning and information-theoretic analysis. We propose an efficient strategy that pre-trains the transmitter in a task-agnostic and label-free manner, followed by joint fine-tuning of both the transmitter and receiver in a task-specific, label-aware manner. Simulation results show that our proposed method reduces training communication overhead to about half that of full-supervised methods using the SGD optimizer, demonstrating significant improvements in training efficiency. Copyright © 2025, The Authors. All rights reserved.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Regression-based Smart Agriculture Prediction 15

Regression-based Smart Agriculture Prediction

引用

15th International Conference on Advances in Computing, Control, and Telecommunication Technologies, ACT 2024

作者： Kumar, Sachin Sharma, Hirdesh Rajput, Sheetal Kumar, Arun Tiwari, Vijay Kumar Affiliated to GGSIP University New Delhi India U.P. Greater Noida India Dept. of Computer Science Bhagwant Institute of Technology U.P Muzaffarnagar India Dept. of CSE İTS Engineering College U.P. Greater Noida India

ISBN: (纸本)9798331300579

Smart agriculture has come to be seen as a potential method for boosting agricultural output and sustainability. The study primarily focuses on the application of regression algorithms for predictive analysis in smart agriculture with reference to crop yield prediction. This study's main objective is to develop a reliable regression model that estimates crop yields based on a number of agricultural and environmental factors. This is done by accumulating previous agricultural information, such as weather patterns, soil properties, irrigation methods, and crop-specific traits, via smart sensors positioned throughout agricultural areas. In the study, a variety of regression approaches are investigated, including support vector regression (SVR), random forest regression, linear regression, polynomial regression, and more. The dataset is analyzed before using these methods to build prediction models. To validate the model's accuracy and generalizability, the dataset is split into training and testing sets, and cross-validation techniques are employed. Regression algorithms are highly reliable at predicting agricultural yields, according to the results. Random forest regression outperforms all other methods due to its capability to handle non-linearity and capture nuanced relationships between input characteristics and crop yields. However, the findings from other regression algorithms are equally promising, suggesting that they may be useful in specific situations. Additionally, the study evaluates how different input variables impact the model's efficacy. It draws attention to crucial elements including temperature, precipitation, soil moisture, and nutrient levels that have a significant impact on crop yields. Farmers may select the ideal crops, irrigation schedules, and fertilizer applications with the aid of an understanding of these connections. © Grenze Scientific Society, 2024.

关键词： Smart agriculture

来源：评论

学校读者我要写书评

暂无评论

Experiments of Crowd Detection for Crowd Digital Twins 22

Experiments of Crowd Detection for Crowd Digital Twins

引用

22nd IEEE Consumer Communications and Networking Conference, CCNC 2025

作者： Chong, Kuan Pok Lai, Chon Hou Ling, Weibo Lu, Zhuoqian Wang, Rui Wang, Ruoqi Yu, Yanjun Testa, Alex Lam, Chan-Tong Tang, Su-Kit Delnevo, Giovanni Casadei, Roberto Girau, Roberto Mirri, Silvia Macao Polytechnic University China Guangdong University of Foreign Studies China University of Bologna Dept. of Computer Science and Engineering Italy Macao Polytechnic University Faculty of Applied Sciences China

ISBN: (纸本)9798331508050

The development of a crowd digital twin offers significant potential for enhancing public safety, urban planning, and event management. A key challenge in creating such a digital twin lies in the efficient and accurate acquisition of crowd-related data, particularly through object detection models deployed on resource-constrained devices. Through a series of experiments, we compare TinyML and Edge approaches in terms of detection accuracy, inferencing time, and resource utilization. Our findings highlight the trade-offs inherent in selecting detection models for crowd digital twin applications, underscoring the importance of aligning model choice with specific deployment needs. © 2025 IEEE.

关键词： Crowdsourcing

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：