检索结果-内蒙古大学图书馆

International Conference on Computing and Networking Technology (ICCNT)

作者： Bhuvi Ghosh Harsh Shetye Atharv Salian Nilesh Marathe Dept. of Computer Science and Engineering (Data Science) Dwarkadas J. Sanghvi College of Engineering Mumbai India

ISBN: (数字)9798350370249

ISBN: (纸本)9798350370270

Deep convolutional neural networks (CNNs), renowned for their high accuracy on clean datasets, remain notably vulnerable to adversarial attacks. This study assesses how the VGG16 and ResNet50 models withstand various white-box attacks when used for traffic sign classification. Despite their impressive standard accuracy, our findings reveal that these accuracies do not predict robustness against attacks. Techniques such as the Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) substantially lowered model accuracy, while DeepFool and Carlini & Wagner (C&W) attacks were even more disruptive. In particular, ResNet50 was more prone to these attacks, with its accuracy dropping to as low as $\mathbf{1 2. 8 5 \%}$ under DeepFool and $\mathbf{1 7. 4 2 \%}$ under C&W at specific confidence levels. Retraining models with adversarial examples did enhance their resilience to certain attacks but at the expense of reduced effectiveness on clean data. These results highlight the critical need for developing more robust defense mechanisms that protect CNNs against a wide array of adversarial threats without compromising performance on standard inputs.

关键词： Training Accuracy Noise Robustness Data models Iterative methods Standards Residual neural networks Glass box Resilience

来源：评论

学校读者我要写书评

暂无评论

CA-CR: Context-Aware Congestion Resolution Method for Automated Guided Vehicles

CA-CR: Context-Aware Congestion Resolution Method for Automa...

引用

Applied Artificial Intelligence (ICAPAI), International Conference on

作者： Yigit Can Dundar Bernt Arild Bremdal Anne Håkansson Randi Karlsen Dept. of Computer Science UiT - The Arctic University of Norway Tromsø Norway Dept. of Computer Science and Computational Engineering UiT - The Arctic University of Norway Narvik Norway

ISBN: (数字)9798350349764

ISBN: (纸本)9798350349771

Modern logistics solutions use Automated Guided Vehicles (AGVs) in order to streamline logistics tasks in various work environments. While handling logistics tasks, AGVs can encounter congestions in the form of traffic or blockages which can cause slow-downs and deadlocks. Congestions can occur randomly and unexpectedly in work environments that are unknown to the AGVs. In unknown environments, access to environmental information is limited. Therefore, avoiding or resolving a congestion is a difficult task for AGVs. In order to compensate for the lack of environmental information, the contextual information of the congestion can be used. This paper presents the Context-Aware Congestion Resolution (CACR) method for AGVs in unknown work environments. The method uses contextual information in the form of number of agents, the vehicles’ current battery levels, task priorities and agent positioning to resolve congestions through a multi-agent negotiation. The negotiation uses the contextual information to find various solutions to random and unpredictable congestion scenarios. The negotiations are aided with reinforcement learning to train AGVs to move towards a collective utility maximizing behavior. To test the AGVs performance, the trained behavior is compared to a baseline, non-trained behavior for evaluation of reinforcement learning’s impact on optimizing the resolution of congestions. Findings indicate that the trained behavior shows better scaling with number of agents, better battery utilization and collective utility compared to baseline.

关键词： Training Remotely guided vehicles Reinforcement learning System recovery Batteries Resource management Proposals

来源：评论

学校读者我要写书评

暂无评论

Empirical Analysis of Classical and Quantum Algorithms for Portfolio Optimization: Enhancing Financial Decision-Making Through Quantum Computing

Empirical Analysis of Classical and Quantum Algorithms for P...

引用

Sustainable Communication Networks and Application (ICSCNA), International Conference on

作者： Krish Valecha Anuradha Yeole Atharv Salian Shruti Mathur Dept. of Computer Science and Engineering (Data Science) Dwarkadas J. Sanghvi College of Engineering Mumbai India

ISBN: (数字)9798331530013

ISBN: (纸本)9798331530020

This research investigates the efficacy of quantum and classical algorithms in the context of portfolio optimization, focusing on a dataset comprising 20 equities from India's National Stock Exchange over a duration of three years. The study employs quantum methodologies such as the Quantum Approxi-mate Optimization Algorithm (QAOA), the Variational Quantum Eigensolver (VQE), and the Exact Minimum Eigen Solver, alongside conventional strategies including Mean-Variance Op-timization, Genetic Algorithms, and Simulated Annealing. The methodology for this study includes data preprocessing, issue designing, and performance evaluation based on metrics such as MSE, RMSE, and Mean Absolute Error. The results obtained show quantum algorithms exhibit better performance over the classical methods regarding prediction accuracy and risk-return profiles, especially the Exact Minimum Eigen hSolver and QAOA. The Exact Minimum Eigen Solver and QAOA had the lowest MSE values (0.006 and 0.005, respectively), whereas the best-performing classical technique, the Genetic Algorithm, had an MSE of 0.0105. These findings imply that quantum algorithms provide significant advantages in portfolio optimization problems, with the potential to revolutionize financial optimization tactics. The research can be further unfolded into avenues for future inquiry, including expansion of the choice of assets, alternative data sources integration, as well as hybrid quantum-classical models to efficiently enhance portfolio management strategies.

关键词： Quantum algorithm Accuracy Soft sensors Simulated annealing Transforms Prediction algorithms Stock markets Optimization Portfolios Genetic algorithms

来源：评论

学校读者我要写书评

暂无评论

A novel CNN method for the accurate spatial data recovery from digital images

引用

Materials Today: Proceedings 2023年 80卷 1706-1712页

作者： Murugan, G. Moyal, Vishal Nandankar, Praful Pandithurai, O. John Pimo, Er.S. St John College Of Engineering and Management Maharashtra Palghar India SVKMs Institute of Technology Maharastra Dhule India Electrical Engineering Dept. Government College of Engineering Nagpur India Computer Science and Engineering Rajalakshmi Institute of Technology Poonamallee Tamilnadu Chennai India Department of CSE St. Xavier's Catholic College of Engineering Chunkankadai Tamil Nadu Nagercoil India

The parsing of floorplans has been an issue for a long time in automated document processing and with algorithmic methods until recent years. This problem has also improved output with the emergence of convolutionary neural networks (CNN). The job here is to obtain spatial and geometrical data from floorplans as accurately as possible. The aim of this project is to extract the most information from a floor plan image around instance segmentation models like Cascade Mask R-CNN. A new style of key point CNN is being implemented to supplement the segmentation to find correct corner positions. Then the resulting segmentation is combined in a post-processing stage. With a mean IoU of 72.7 percent versus 57.5 percent, the resulting segmentation scores surpass the existing baseline of the CubiCasa5k floorplan data base. Moreover, for almost every class, the mean IoU for each class is increased. Cascade Mask R-CNN has also been shown to be better suited to this role than Mask R-CNN. © 2021

关键词： E-learning

来源：评论

学校读者我要写书评

暂无评论

Real-Time Injury Risk Assessment in Athletes Based on Relative Joint Angles

Real-Time Injury Risk Assessment in Athletes Based on Relati...

引用

Emerging Smart Computing and Informatics (ESCI), Conference on

作者： Atharv Salian Bhuvi Ghosh Harsh Shetye Parth Kapadia Kriti Srivastava Dept. of Computer Science and Engineering (Data Science) Dwarkadas J. Sanghvi College of Engineering Mumbai India

Preventing running injuries is critical for track and field athletes. This work presents a novel machine learning methodology to automatically detect injury risk from videos by analyzing biomechanical technique. Joint angles from anatomical pose estimates serve as descriptive features input to ML classifiers. Statistical tests verified significant distribution differences in angles between injured and non-injured states, evidencing utility as injury markers. Multi-frame analysis integrating temporal context attained 0.735 accuracy on held-out test data using SVM, surpassing single-frame approaches. The scoring function showed no significant statistical drift from training to test, providing evidence of strong generalization. By automating nuanced pose pattern quantification imperceptible to the naked eye, the proposed video screening enables early at-risk athlete identification for targeted preventative intervention before injuries develop.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Blockchain Based Verification of Vehicle History for Pre-owned Vehicle Industry

Blockchain Based Verification of Vehicle History for Pre-own...

引用

2021 International Conference on Communication, Control and Information sciences, ICCISc 2021

作者： Vishnu Gopal, P.G. Mathew, George Government Engineering College Dept. of Computer Science and Engineering Thrissur India Government Engineering College Dept. of Computer Science and Engineering Thrissur India

ISBN: (纸本)9781665412797

The automotive industry is one of the lucrative markets in the world. With the advancement of the human lifestyle, everybody needs a car for their daily business. Before buying a used vehicle, the history of the vehicle should be verified. Currently, the history of the vehicle is obtained from the second-hand vehicle dealers. The data provided by dealers need not be correct because they do not maintain a proper record. The only way for the buyer is to believe the dealer and proceed with the deal. To resolve these problems, we proposed a distributed framework using blockchain technology to verify the history of the vehicles. The creator or deployer in the system is responsible for creating the registration of the vehicle with a unique VIN and it will be updated to the shared ledger. Then the vehicles are distributed to the trusted dealers in the system with the execution of the smart contract in the supply chain and finally released to the user. In the running phase, the proposed model records insurance, real-Time data and maintenance services, and legal issues if any. The main advantage of the vehicle verification on the blockchain is that it can reduce the possibility of fraud and counterfeiting documents. © 2021 IEEE.

关键词： Smart contract

来源：评论

学校读者我要写书评

暂无评论

IoT-Enabled Covert Surveillance Camera Identification through IR Sensor

IoT-Enabled Covert Surveillance Camera Identification throug...

引用

Advancement in Computation & computer Technologies (InCACCT), International Conference on

作者： Shaik Ruksana Shaik Imran Shaik Fatima Farheen Krovi Mohan Kireeti K L Sailaja Dept. of Artificial Intelligence and Data Science VR Siddhartha Engineering College Vijayawada India Dept. of Computer Science and Engineering VR Siddhartha Engineering College Vijayawada India

ISBN: (数字)9798350371314

ISBN: (纸本)9798350371321

A rising number of spy cameras installed in sensitive locations globally poses an escalating threat to personal privacy. Detecting these small and well-concealed cameras with the naked eye is challenging. Despite the abundance of Android apps, accurate detection remains elusive, with many requiring subscriptions and some charging users without prior notice. Furthermore, most IoT devices are designed to detect wired cameras, overlooking the prevalence of wireless spy cameras. Existing studies often rely on impractical methods, such as wireless transmission detection. The primary objective of this paper is to develop a cost-effective, scalable, robust, and user-friendly spy camera detector. This detector identifies all night vision cameras emitting infrared light, promptly alerting the user through a beep sound from the attached buzzer. In addressing the imperative need for heightened privacy protection, the IoT spy camera detector adept.y spots concealed cameras via IR signals. By offering user-friendly location alerts through a reassuring buzzer, it actively contributes to enhancing security measures against evolving privacy threats, creating a safer environment for all.

关键词： Wireless communication Night vision Privacy Wireless sensor networks Surveillance Detectors Cameras

来源：评论

学校读者我要写书评

暂无评论

Optimizing Fashion Recommendations for Diverse Body Types: An Epsilon Greedy Approach

Optimizing Fashion Recommendations for Diverse Body Types: A...

引用

International Conference on Computing and Networking Technology (ICCNT)

作者： Anuradha Yeole Ami Desai Kriti Srivastava Dept. of Computer Science and Engineering (Data Science) Dwarkadas J. Sanghvi College of Engineering Mumbai India

ISBN: (数字)9798350370249

ISBN: (纸本)9798350370270

As the fashion industry evolves, the need for sophisticated recommendation systems on e-commerce platforms becomes increasingly critical. This paper introduces a novel reinforcement learning-based recommendation system tailored for diverse body types using the Epsilon Greedy algorithm. Our dual-layered model architecture integrates content-based filtering to precisely identify user body types through innovative physical measurements and photographic analysis. This groundwork allows for a dynamic reinforcement learning strategy to refine and personalize fashion recommendations. Statistical analysis confirmed that our system significantly enhances recommendation precision, achieving a novelty score of $\mathbf{84\%}$ and a click-through rate of 46%, surpassing traditional static models. The adaptive reinforcement mechanism demonstrated robust generalization across various user interactions without significant deviation in performance. By personalizing fashion recommendations to individual user preferences and body types, our system facilitates more accurate and engaging shopping experiences, effectively addressing the challenges of personalization in the fast-evolving e-commerce sector.

关键词： Industries Data privacy Analytical models Filtering Statistical analysis Reinforcement learning Market research Electronic commerce Object recognition Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Acoustic Scene Classification using Fusion of Features and Random Forest Classifier 9

Acoustic Scene Classification using Fusion of Features and R...

引用

9th International Conference on Computing for Sustainable Global Development, INDIACom 2022

作者： Sachdeva, Shantanu Mulimani, Manjunath Manipal Institute of Technology Manipal Academy of Higher Education Dept. of Computer Science and Engineering Manipal India

ISBN: (纸本)9789380544441

This paper proposes a model for the task of Acoustic Scene Classification. The proposed model utilizes convolutional neural networks and a random forest classifier to predict the class of the audio clips. The features used by the proposed model are log-Mel, Mel-frequency cepstral coefficient, and Gammatone cepstral coefficient spectrograms. Each spectrogram is processed using a convolutional neural network and combined into a single vector. The processed feature vectors are classified into one of the acoustic scenes using the random forest classifier. The proposed model is evaluated on Tampere University of Technology Urban Acoustic Scenes 2018 and the Tampere University Urban Acoustic Scenes 2019 development datasets. The performance of the proposed model is compared with the Detection and Classification of Acoustic Scenes and Events 2018 and 2019 challenge baseline models to show its efficacy. The proposed model has an accuracy of 68.1% and 67.1% for the two datasets, respectively. © 2022 Bharati Vidyapeeth, New Delhi.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

ChatGPT or A Silent Everywhere Helper: A Survey of Large Language Models

arXiv

引用

arXiv 2025年

作者： Akhtarshenas, Azim Dini, Afshin Ayoobi, Navid dept. of Telecommunication Engineering Universitat Politècnica de València Valencia Spain dept. of Computer Science Tampere University Tampere Finland dept. of Computer Science University of Houston Houston United States

Large Language Models (LLMs) have revolutionized natural language processing Natural Language Processing (NLP), with Chat Generative Pre-trained Transformer (ChatGPT) standing out as a notable example due to its advanced capabilities and widespread applications. This survey provides a comprehensive analysis of ChatGPT, exploring its architecture, training processes, and functionalities. We examine its integration into various domains across industries such as customer service, education, healthcare, and entertainment. A comparative analysis with other LLMs highlights ChatGPT’s unique features and performance metrics. Regarding benchmarks, the paper examines ChatGPT’s comparative performance against other LLMs and discusses potential risks such as misinformation, bias, and data privacy concerns. Additionally, we offer a number of figures and tables that outline the backdrop of the discussion, the main ideas of the article, the numerous LLM models, a thorough list of datasets used for pre-training, fine-tuning, and evaluation, as well as particular LLM applications with pertinent references. Finally, we identify future research directions and technological advancements, underscoring the evolving landscape of LLMs and their profound impact on artificial intelligence Artificial Intelligence (AI) and society. © 2025, CC BY-NC-ND.

关键词： Natural language processing systems

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：