检索结果-内蒙古大学图书馆

28th International Conference on Knowledge Based and Intelligent information and Engineering Systems, KES 2024

作者： Akber, Syed Muhammad Abrar Szczesna, Agnieszka Kazmi, Sadia Nishat Department of Computer Graphics Vision and Digital Systems Faculty of Automatic Control Electronics and Computer Science Silesian University of Technology Gliwice44-100 Poland Faculty of Automatic Control Electronics and Computer Science Silesian University of Technology Gliwice44-100 Poland

Longer training times pose a significant challenge in artificial neural networks (ANNs) as it may leads to increasing the computational costs and decreasing the effectiveness of the model. Therefore, it is imperative to reduce training times in ANNs to enhance the computational efficiency. The initialization of the weights between the layers in ANN plays a vital role in reducing training times. Appropriate weight initialization can help the network converge faster during the training by providing an optimum starting point for the network. Therefore, weight initialization techniques are essential for efficient training of ANNs. This paper revisits and implements different popular weight initialization techniques in ANNs and analyzes their impact on training time. Specifically, this paper implements Gaussian-based, Kaming-based, and Xavier-based weight initiation atop a popular DNN-based network. The experiments are conducted by employing a well-known dataset. The results show that the scenario when no weight initiation is applied consumed the highest training time, whereas different weight initiation techniques contribute in reducing the training times for the network. © 2024 The Authors.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Bias Mitigation in Medical Datasets: Enhancing GBM and KNN Predictive Models with SMOTE Techniques 2

Bias Mitigation in Medical Datasets: Enhancing GBM and KNN P...

引用

2nd International Conference on Computing and Data Analytics, ICCDA 2024

作者： Boughareb, Djalila Bordjiba, Hana Boughareb, Rima Seridi, Hamid University of 8 Mail 945 Guelma Labstic Laboratory Computer Science Department Guelma Algeria University of 8 Mai 1945 Guelma Computer Science Department Guelma Algeria University of Badji Mokhtar Computer Science Department Annaba Algeria

ISBN: (纸本)9798331542573

Dataset bias is a critical issue in medical machine learning models that can significantly impact the efficacy of disease prediction and diminish the ability to generate relevant clinical decisions. To address this problem, oversampling the underrepresented class presents a promising solution. This study investigates the impact of oversampling techniques, specifically SVM-SMOTE and BorderlineSMOTE, on both machine learning using K-Nearest Neighbors (KNN) algorithm and Ensemble Learning using Gradient Boosting Machine (GBM). we assessed changes in accuracy, precision, recall, F1 score, Positive Predictive Value (PPV), Equal Opportunity Difference (EOD), Disparate Impact (DI), and Imbalance Ratio (IR) across two medical datasets. In heart disease prediction using Gradient Boosting Machine (GBM), both SVM-SMOTE and BorderlineSMOTE significantly enhanced key metrics. BorderlineSMOTE maintained high accuracy at 0.85, with a notable increase in recall to 0.92., indicating improved identification of positive cases. Precision and F1 score also showed robust improvements, reflecting the models' enhanced ability to accurately classify instances across all classes. Also, the introduction of synthetic samples notably reduced EOD and DI values, indicating a reduction in bias within model predictions. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Log Analysis for Feature Engineering and Application of a Boosting Algorithm to Detect Insider Threats 3rd

Log Analysis for Feature Engineering and Application of a Bo...

引用

3rd International Conference on Intelligent Systems and Pattern Recognition, ISPR 2023

作者： Besnaci, Samiha Hafidi, Mohamed Lamia, Mahnane LRS Laboratory computer science department Badji Mokhtar University Annaba Algeria

ISBN: (纸本)9783031463341

The insider threat has captured the attention of a large number of researchers, as a sensitive and critical issue for most organizations in today’s digital world. It is also a major source of information security and can cause more damage and financial loss than any other threat. In this article, we’ve used feature engineering for features that represent users’ day-to-day activities. We tried different machine learning models such as random forest, xgboost and Catboost. Since the data used to detect malicious activity is unbalanced, the target audience is small. We used KMeansSmote to balance the classes of learning so that the algorithms can learn both classes well. And we used the catboost algorithm to identify the malicious user. The dataset used to evaluate this model is Cert v4.2. CatBoost outperformed other models with the highest F1-score of 95%. © 2024, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Lax-net: Freezing of Gait Detection in Parkinson's Disease Using LSTM with Attention and XGBoost

Lax-net: Freezing of Gait Detection in Parkinson's Disease U...

引用

2023 International Conference on Decision Aid sciences and Applications, DASA 2023

作者： Ayoub, Bouslah Aicha, Khaoucha Nora, Taleb Bouthaina, Messaoud Badji Mokhtar University Lri Laboratory Computer Science Department Annaba Algeria Computer Science Department Annaba Algeria

ISBN: (纸本)9798350342055

Parkinson's disease (PD) is a neurodegenerative disorder marked by motor symptoms like gait freezing (FoG), which notably affects patients' mobility and overall quality of life. Early detection of FOG episodes is crucial for timely intervention and reducing the risk of falls and injuries. In this paper, we present Lax-net, a deep learning-based approach for the early detection of FOG using wearable sensor data with time series information. These data often exhibit complex temporal dependencies, making it challenging to extract meaningful features and accurately classify them. To address this challenge, our proposed method combines the strengths of LSTM with attention mechanisms to capture temporal relationships and identify important features in the data. The derived features are then fed into the XGBoost model, a robust classifier with the ability to grasp non-linear relationships. We evaluate the performance of our proposed method on a benchmark dataset, comparing it with other state-of-the-art methods. The results demonstrate that our combined approach achieves superior accuracy and outperforms individual models such as LSTM and XGBoost. © 2023 IEEE.

关键词： Wearable sensors

来源：评论

学校读者我要写书评

暂无评论

Reaching and Grasping with NAO Robot 16th

Reaching and Grasping with NAO Robot

引用

16th International Conference on Social Robotics, ICSR + AI 2024

作者： Borghese, N. Alberto Chitti, Eleonora Laboratory of Applied Intelligent Systems Department of Computer Science University of Milan Milan Italy

ISBN: (纸本)9789819635184

In this paper we propose a robust approach that combines classical computer vision with a smart setup that enables a NAO humanoid robot to play with children. Blocks are adopted as playing objects as they are widely used and are a highly versatile support. This approach addresses the challenges of object reaching and grasping, by allowing the robot to approach the table over which activity is carried out, orient appropriately, plan the reaching movement that is executed partly through direct and partly through inverse kinematics, grasp the block, moving it over the table and releasing it into its target position. By integrating visual information provided by the robot cameras into a closed-loop control system, no additional sensors are required. This approach enables not only children to play with a Humanoid robot in a richer way, but also to be used to treat cognitive / motor deficits inside what is termed Robot Assisted Therapy. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Anthropomorphic robots

来源：评论

学校读者我要写书评

暂无评论

Human Activity Recognition in Enhancing Healthcare for Aging Populations: Challenges, Innovations, and Future Directions 1

Human Activity Recognition in Enhancing Healthcare for Aging...

引用

1st International Conference on Innovative and Intelligent Information Technologies, IC3IT 2024

作者： Boulahia, Mohamed Abderrahmen Benharzallah, Saber Titouna, Faiza Dilekh, Tahar University of Batna 2 Lamie Laboratory Department of Computer Science Batna Algeria

ISBN: (纸本)9798331541064

As the global population ages, healthcare and social systems face escalating demands. Effective solutions for promoting healthy aging increasingly rely on deep learning and Internet of Things technologies to provide personalized, preventive, and proactive care. Central to these efforts is human activity recognition, which enables the continuous monitoring and assessment of daily activities - a vital indicator of well-being and independence in older adults. This paper highlights key concepts and methodologies in applying deep learning and internet of things in human activity recognition, emphasizing its role in understanding and enhancing the quality of life for the elderly through accurate, activity-based health insights. However, most current studies continue to emphasize achieving high accuracy in machine learning models, despite already reaching substantial levels of precision. This paper argues for the need to shift focus towards developing adaptable models capable of recognizing new and varied activities. Additionally, it highlights the importance of recommendation systems in healthcare, which should extend beyond mere activity recognition to provide actionable advice and preventive measures for the elderly. This study also addresses the gap in the literature by providing a comprehensive review of existing approaches and identifying key research directions that should be prioritized in the field. By signaling the lack of comprehensive studies offering in-depth insights like this paper, it underscores the necessity of moving beyond current achievements towards more dynamic and context-aware solutions in human activity recognition technologies. © 2024 IEEE.

关键词： Active learning

来源：评论

学校读者我要写书评

暂无评论

An Empirical Study on Automated Test Generation Tools for Java:Effectiveness and Challenges

引用

Journal of computer science & Technology 2024年第3期39卷 715-736页

作者：刘相君余萍马晓星 State Key Laboratory for Novel Software Technology Nanjing UniversityNanjing 210023China Department of Computer Science and Technology Nanjing UniversityNanjing 210023China CCF ACM IEEE

Automated test generation tools enable test automation and further alleviate the low efficiency caused by writing hand-crafted test ***,existing automated tools are not mature enough to be widely used by software testing *** paper conducts an empirical study on the state-of-the-art automated tools for Java,i.e.,EvoSuite,Randoop,JDoop,JTeXpert,T3,and *** design a test workflow to facilitate the process,which can automatically run tools for test generation,collect data,and evaluate various ***,we conduct empirical analysis on these six tools and their related techniques from different aspects,i.e.,code coverage,mutation score,test suite size,readability,and real fault detection *** discuss about the benefits and drawbacks of hybrid techniques based on experimental ***,we introduce our experience in setting up and executing these tools,and summarize their usability and ***,we give some insights into automated tools in terms of test suite readability improvement,meaningful assertion generation,test suite reduction for random testing tools,and symbolic execution integration.

关键词： automated test generation search-based software testing random testing symbolic execution

来源：评论

学校读者我要写书评

暂无评论

Parameter-free Zero-shot 3D Model Classification Based on Multi-view Representation 24

Parameter-free Zero-shot 3D Model Classification Based on Mu...

引用

2024 International Conference on Intelligent Perception and Pattern Recognition, IPPR 2024

作者： Xi, Shuting Bai, Jing Yan, Hao School of Computer Science and Engineering North Minzu University China The Key Laboratory of Images Graphics Intelligent Processing of State Ethnic Affairs Commission China

ISBN: (纸本)9798400707315

The large-scale pre-trained models have demonstrated excellent performance in processing visual and language tasks in open-world scenarios. However, recent studies have shown that there are limitations in using comparative language-vision pretrained models for zero-shot 3D model classification, mainly in ignoring the multi-view independent information of 3D models, and using only a single template for text prompts, which fails to exploit the rich contextual information provided by the language, and restricts the matching of visual and semantic information. In order to solve these problems, this paper proposes a new framework of Parameter-free zero-shot 3D model classification based on "interaction first, fusion later", which utilizes a pre-trained visual encoder to extract multi-view visual features, and then explores contextual information from multiple perspectives with the help of semantic comprehension of a large language model, establishes fine-grained associations between independent views and prompts through view-by-view interaction with text features, and integrates the independent decision results among views through decision-level fusion strategy to realize zero-shot classification. Without any 3D training, the proposed method achieves 78.8% accuracy on the ZS3D dataset, and 53.2%, 40.3%, and 26.3% accuracy on the three sub-datasets of Ali, which verifies the generalizability and effectiveness of the proposed algorithm. © 2024 Copyright held by the owner/author(s).

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

ARnopoly: exploring strengths and weaknesses of AR experience enhancing board games 3

ARnopoly: exploring strengths and weaknesses of AR experienc...

引用

3rd Workshop on Artificial Intelligence for Human-Machine Interaction, AIxHMI 2024

作者： Chitti, Eleonora Barbagallo, Andrea Delia, Andrea Borghese, N. Alberto Applied Intelligent Systems Laboratory Department of Computer Science University of Milan Italy

This project aims to explore and apply the best practices of Augmented Reality (AR) design to the board play. We adapted the Monopoly game, developing an AR version that integrates the traditional game with new elements to take advantage of the virtual world. Our goal is to explore which AR game rules and mechanics can be incorporated into board game design to create a more engaging and effective experience for players. With ARnopoly, we specifically examined the strengths of augmented reality and the weaknesses that need to be addressed to seamlessly integrate the physical board play with the digital experience. ARnopoly was tested involving 7 participants, and from the analysis of their feedback it emerged that the integration of the board play with AR increases the immersion in the game and the overall positive feeling. However, in the long run, since the monopoly-like game is very long-lasting, this positivity is dampened by the constraint of holding the device in hand which causes fatigue in the user. © 2024 for this paper by its authors.

关键词： Game design

来源：评论

学校读者我要写书评

暂无评论

A robust approach to authorship verification using siamese deep learning: application in phishing email detection

引用

International Journal of Speech Technology 2024年第2期27卷 405-412页

作者： Remmide, Mohamed Abdelkarim Boumahdi, Fatima Ammar Aouchiche, Imane Rebeh Guendouz, Amina Boustia, Narhimene LRDSI laboratory Department of Computer Science Faculty of Sciences University of Blida 1 Blida Algeria LRDSI laboratory Faculty of Technology University of Blida 1 Blida Algeria

Given the rapid and significant increase in email data, it is crucial for both individuals and organisations to prioritise the implementation of strong cybersecurity measures to combat attacks such as phishing emails. While continuous research has been made to find the most efficient approach to combating phishing emails, cybercriminals on the other hand continue to be determined, and their techniques become more and more sophisticated. In this paper, we present a novel approach for email classification using a Siamese deep learning network in order to verify authorship. We used the writing style and behavioural characteristics of the authors in this approach to identify the true source of emails and classify them as phishing, legitimate, harassment, or suspicious. This model was evaluated on the SeFACED dataset, where it received an accuracy of 90,12%, showcasing its efficacy in classifying emails, which enhanced email security and contributed to a safer online environment. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Electronic mail

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：