检索结果-内蒙古大学图书馆

12th International Symposium on Digital Forensics and Security, ISDFS 2024

作者： Jahanzeb Khan, M. Rath, Suman Zaib, Muhammad Hassan University of Nevada Reno Dept. of Computer Science and Engineering Reno United States Air University Dept. of Computer Science Islamabad Pakistan

ISBN: (纸本)9798350330366

This research addresses the dual challenges of image restoration quality and data privacy in optical remote sensing. Traditional restoration methods often fall short due to the complex nature of remote sensing images, and data privacy concerns further complicate the use of advanced techniques. By integrating the Deep Memory Connected Neural Network (DMCN) with the Data-Decoupled Federated Learning (DDFL) framework, our approach enables significant improvements in image restoration without requiring direct access to sensitive raw data. This method not only enhances data privacy by leveraging federated learning principles but also incorporates advanced techniques like Gaussian image denoising to maintain high restoration quality despite potential noise introduced by the federated process. The performance of the federated DMCN, particularly on the UCMERCED dataset, demonstrates minimal accuracy degradation, even in the presence of noise, while the strategic use of Downsampling Units within DMCN optimizes computational efficiency. Our comprehensive evaluations reveal the effectiveness of this approach in balancing data privacy with the need for high-quality image restoration, suggesting a promising direction for future advancements in remote sensing applications. © 2024 IEEE.

关键词： Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Utilization of Decentralized Finance (DeFi) and Distributed Ledger Technology (DLT) in Banking operations

Utilization of Decentralized Finance (DeFi) and Distributed ...

引用

2024 International Conference on Intelligent Computing and Sustainable Innovations in Technology, IC-SIT 2024

作者： Kumar, Ch Sree Singh, Akhilendra Pratap Reddy, K Hemant K National Institute of Technology Meghalaya Dept. of Computer Science Meghalaya India VIT-AP Dept. of Computer Science and Engineering Vijayawada India

ISBN: (纸本)9798350369175

Distributed Ledger Technology (DLT) is a decentralized database system where transactions are recorded and verified across multiple nodes. Its key features include immutability, time-stamping, and consensus-based validation. Numerous DLT applications are in supply chain management, intellectual property, cross border payments, energy trading, real estate, and online donations. DeFi, a combination of cryptocurrency and blockchain technology, offers financial services without intermediaries. Hence transactions with various digital assets based on cryptocurrency price feeds needs an automated framework. Smart contracts, self-executing contracts with terms directly written into code, automate processes and reduce manual intervention. This paper proposes a decentralized financial trading model using the AAVE protocol. AAVE is a decentralized platform that allows for transactions with various digital assets based on cryptocurrency price feeds. Proposed model uses Chainlink, a decentralized oracle network, to provide accurate and reliable price feeds. IPFS is used for data storage, while Graph is employed for indexing and querying blockchain data. The paper presents an example of a DeFi protocol to simulate banking operations, showcasing the potential of DeFi and DLT in revolutionizing traditional banking processes. © 2024 IEEE.

关键词： Distributed ledger

来源：评论

学校读者我要写书评

暂无评论

EMPIRICAL ANALYSIS OF IEEE754, FIXED-POINT AND POSIT IN LOW PRECISION MACHINE LEARNING

UPB Scientific Bulletin, Series C: Electrical Engineering an...

引用

UPB Scientific Bulletin, Series C: Electrical engineering and computer science 2023年第3期85卷 13-24页

作者： Ciocîrlan, Ștefan-Dan Neacșu, Teodor-Andrei Rughiniș, Răzvan-Victor University POLITEHNICA of Bucharest Dept. of Computer Science Romania

Deep neural networks have changed the current algorithms’ results in applications such as object classification, image segmentation or natural language processing. To increase their accuracy, they became more complex and more costly in terms of storage, computation time and en-ergy consumption. This paper attacks the problem of storage and presents the advantages of using different number representations as fixed-point and posit numbers for deep neural network inference. The deep neural networks were trained using the proposed framework Low Precision Machine Learning (LPML) with 32-bit IEEE754. The storage was first optimized by the usage of knowledge distillation and then by modifying layer by layer the number representation together with the precision. The first significant results were made by modifying the number representation of the network but keeping the same precision per layer. For a 2-layer network (2LayerNet) using 16-bit Posit, the accuracy is 93.45%, close to 93.47%, the accuracy for using 32-bit IEEE754. Using the 8-bit Posit decreases the accuracy by 1.29%, but at the same time, it reduces the storage space by 75%. The usage of fixed point representation showed a small tolerance in the number of bits used for the fractional part. Using a 4-4 bit fixed point (4 bits for the integer part and 4 bits for the fractional part) reduces the storage used by 75% but decreases accuracy as low as 67.21%. When at least 8 bits are used for the fractional part, the results are similar to the 32-bit IEEE754. To increase accuracy before reducing precision, knowledge distillation was used. A ResNet18 network gained an 0.87% increase in accuracy by using a ResNet34 as a professor. By changing the number representation system and precision per layer, the storage was reduced by 43.47%, and the accuracy decreased by 0.26%. In conclusion, with the usage of knowledge distillation and change of number representation and precision per layer, the Resnet18 network had 66.75% sm

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Breast Cancer Classification via the Use of ML and DL: A Comprehensive Review 3

Breast Cancer Classification via the Use of ML and DL: A Com...

引用

3rd International Conference on Automation, Computing and Renewable Systems, ICACRS 2024

作者： Sarang, Ruvanshi Dabhi, Vipul Shah, Arpit Dept. of Computer Science Engineering Parul University Vadodara India Dept. of Computer Engineering Parul University Vadodara India

ISBN: (纸本)9798331532420

New and improved methods of diagnosis are needed because breast cancer is still the leading cancer-related killer worldwide. Updates to the methods used to categorize breast cancer have emerged because of recent advances in DL and ML. This review paper’s goal in conducting this research is to bring together existing breast cancer diagnostic and classification methods that make use of deep learning and machine learning techniques. We examine the many approaches, datasets, and performance indicators used in the industry and point out the pros and cons of each as part of the review paper. This review aims to thoroughly examine all the methods that are now in use, spot any trends, and propose new areas for research. This review helps researchers and practitioners understand the current state of breast cancer categorization by combining the findings of past research. Discuss recent advances in ML/DL techniques for improving classification, such as ensemble methods, transfer learning, or hybrid models. Additionally, propose the use of fine-tuning of hyperparameters or novel architectures to enhance classification. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

DURATION-BASED INVESTIGATION OF USER CONTENT CHOICES IN THE EXIT OF FILTER BUBBLES 6

DURATION-BASED INVESTIGATION OF USER CONTENT CHOICES IN THE ...

引用

6th International Congress on Human-computer Interaction, Optimization and Robotic Applications, HORA 2024

作者： Bunza, Rabiatu DEPT. OF COMPUTER SCIENCE FACULTY OF SCIENCE AND ENGINEERING SWANSEA UNIVERSITY United Kingdom

ISBN: (纸本)9798350394634

Content personalization on social platforms has been linked to the creation of filter bubbles. The algorithms provide content recommendations based on the user’s browsing history and interests, which limits content diversity. Studies have been conducted on platforms such as Facebook and YouTube to investigate user awareness and strategies to exit filter bubbles. However, no study has employed a multi-platform, user-centered perspective in investigating filter bubble exit through change in search content, as well as the duration of the content transition across multiple platforms. This study examines the duration of user exit of filter bubbles, across three social platforms;TikTok, YouTube Shots, and Instagram Reels, as well as the impact of word similarity on the duration of the transition between contents, across the platforms. The methodology applied involved triggering the filter bubble effect through user interactions with the search content across the different platforms. Findings revealed that the mean time of a transition between search terms on TikTok is significantly higher than the mean time on Instagram reels, and YouTube shorts. Similarly, the study identified differences in the duration of transition across the different platforms, due to the similarity of the search terms. Apart from improving user awareness of a filter bubble exit strategy, this study contributes to Human-computer Interaction research by highlighting some design implications for future social media application design. © 2024 IEEE.

关键词： Social networking (online)

来源：评论

学校读者我要写书评

暂无评论

Single Object Tracking in Satellite Videos Through Deep Learning and Motion Detection Techniques 6

Single Object Tracking in Satellite Videos Through Deep Lear...

引用

6th International Congress on Human-computer Interaction, Optimization and Robotic Applications, HORA 2024

作者： Hayder, Israa Alizadeh, Fattah Dept. of Computer Science and Engineering University of Kurdistan Hawlere Erbil Iraq Dept. of computer science Toronto Metropolitan university Toronto Canada

ISBN: (纸本)9798350394634

Single object tracking in satellite videos has recently gained a lot of attention in the field of computer vision. Although the satellite images are very informative, the small size of the objects and limited spatial and temporal resolution might make object tracking difficult and inaccurate. In this paper, we propose a method for accurately tracking small objects in satellite videos. In the initial phase, we applied the Fast Super-Resolution Convolutional Neural Network (FSRCNN), a deep learning-based super-resolution technique, to enhance the resolution of the frames. This procedure serves to sharpen the details and features of the objects, resulting in a clearer representation. The main contribution, however, lies in our incorporation of a robust motion detection technique into our framework, marking a significant step forward in this study. The motion detection approach combines various methods and operations with the primary goal of making moving objects distinctly visible and easily distinguishable from the complex image background. This enhancement significantly contributes to the precision and accuracy of the tracking system. We combined both approaches with Fully Convolutional Siamese Network (SiamFC) due to its robustness in tracking processes across frames in videos, capacity to handle changes in target appearance, and ability to deal with small datasets. The Satellite Video Single Object Tracking (SatSOT) and SV248A10-SOT (SV248S), two satellite video single object tracking datasets, were used to evaluate the proposed approach. When the techniques were applied, the SatSOT dataset showed an 11% precision score increase and a 10% success rate improvement. The SV248S dataset exhibited even more significant improvements, with precision scores increasing by 18% and success scores rising by 9%. © 2024 IEEE.

关键词： Satellites

来源：评论

学校读者我要写书评

暂无评论

A Multistage Cascaded Neural Network Model to Identify the Severity of Epileptic Attack

A Multistage Cascaded Neural Network Model to Identify the S...

引用

2024 IEEE International Conference on Signal Processing, Informatics, Communication and Energy Systems, SPICES 2024

作者： Simon, Martha Thampi, Lidiya Lilly Dept. of Computer Science College of Engineering Trivandrum India Indian Institute of Information Technology Kottayam Dept. of Computer Science Kottayam India

ISBN: (纸本)9798350376135

Epilepsy is a neurological disorder characterized by recurrent seizures, impacting the quality of life for millions of individuals worldwide. Timely and accurate detection of epileptic seizures is crucial for proper diagnosis and effective treatment. Electroencephalogram (EEG) signals provide vital information about brain activity and serve as a valuable tool for epilepsy detection. The methodology opted for this study was an enhanced multilayer feed forward neural network for automatic detection of epilepsy from EEG signals and hence predicting the severity. The authors build a multistage cascaded neural network model for achieving the same. Initially a Simple Artificial Neural Network (SimpleANN) is designed to classify EEG segments into seizure, healthy and interictal. Later a Multilayer Feed Forward Neural Network (MFFNN) is employed for detecting the severity of epileptic patients and hence can be classified into primary and secondary stages. The experimental results showed that the proposed SimpleANN classifier had provided noticeable results, and outperforms all tested classifiers. The study involves the collection of EEG data from a diverse group of patients, including those with epilepsy and healthy individuals. This automated epilepsy identifying system passes through different stages including decomposition, feature extraction for achieving a better performance model. The proposed system was tested and compared with SVM, KNN, MLP and Decision Tree under different measurement metrics. ANN has managed to give an accuracy of 80 - 90% which is significantly higher as compared to other machine learning models. The proposed work can automatically learn complex patterns from EEG data provides an advantage in identifying subtle seizure related features that may not be evident through manual analysis. Furthermore, the model offers real-time capabilities, making it suitable for continuous monitoring of patients at risk of seizures. © 2024 IEEE.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

ANALYSIS OF SCIENTIFIC COMPUTING ALGORITHMS UNDER DIFFERENT NUMBER REPRESENTATION SYSTEMS

UPB Scientific Bulletin, Series C: Electrical Engineering an...

引用

UPB Scientific Bulletin, Series C: Electrical engineering and computer science 2023年第1期85卷 29-40页

作者： Ciocîrlan, Ștefan-Dan Soare, Radu-Georgian Țăpuș, Nicolae Resul, Ebru Rughiniș, Răzvan-Victor University POLITEHNICA of Bucharest Dept. of Computer Science Romania

Scientific computing-based applications need a high load of computations. These computations are closely linked to the number representation system (NRS). A benchmark for decimal accuracy, storage space, computation time and energy consumption using multiple NRSs is needed for the base scientific computing algorithms. Eleven NRSs were evaluated under four types of benchmarks: matrix multiplication, solving a linear system of equations using the gradient conjugate method, integral calculation using Simpson’s formula and N-body simulation. The results validate the posit "golden zone" and offer a perspective on different NRSs accuracy related to input ranges. IEEE754 performed well regarding decimal accuracy in all four benchmarks (top 3). © 2023, Politechnica University of Bucharest. All rights reserved.

关键词： Conjugate gradient method

来源：评论

学校读者我要写书评

暂无评论

Comprehensive Women's Health Risk Assessment Platform 2

Comprehensive Women's Health Risk Assessment Platform

引用

2nd International Conference on Artificial Intelligence and Machine Learning Applications, AIMLA 2024

作者： Navya, S. Saha, Sharmistha Nagarajan, Deepa Sathyabama Institute of Science and Technology Dept. of Computer Science and Engineering Chennai India

ISBN: (纸本)9798350349221

The goal of this research is to develop a machine learning-based method for predicting several diseases in women, with an emphasis on polycystic ovarian syndrome (PCOs), diabetes, heart disease, and breast cancer. Timely and accurate prediction of these disorders is essential for early intervention and prevention due to their impact on women's health and prevalence. Predictive models are trained using a wide variety of medical data, including genetic markers, lifestyle factors, demographic data, and clinical test findings. To find patterns and connections between these predictors and the incidence of diseases, machine learning methods are used, such as decision trees, random forests, and support vector machines. The models are able to forecast with accuracy the probability that a woman would develop one or more of these diseases by efficiently obtaining and evaluating data from these datasets. Metrics including accuracy, precision, recall, and F1-score are used to assess the models' performance;the outcomes demonstrate the models' promising levels of predictive power. The creation of a system like this can facilitate targeted screening and preventive measures, assist medical practitioners in identifying high-risk individuals, and eventually lead to better patient outcomes. © 2024 IEEE.

关键词： Diseases

来源：评论

学校读者我要写书评

暂无评论

A Number of Conceptual Scalable Node-Organizing Multi-Tiered Blockchain Architectures for IoT 17

A Number of Conceptual Scalable Node-Organizing Multi-Tiered...

引用

IEEE Congress on Cybermatics: 17th IEEE International Conference on Internet of Things, iThings 2024, 20th IEEE International Conference on Green Computing and Communications, GreenCom 2024, 17th IEEE International Conference on Cyber, Physical and Social Computing, CPSCom 2024, 10th IEEE International Conference on Smart Data, SmartData 2024

作者： Elsaadany, Riham Bégin, Guy Computer Science Dept. Canada

ISBN: (纸本)9798350351637

The high volume and rapid pace of transactions generated by IoT devices pose challenges for current blockchain designs, which typically employ flat or two-tiered node organizations. These models often lack the scalability, throughput, and consensus efficiency required for the IoT environment. This study proposes a multi-tiered clustered design concept that enhances blockchain scalability and efficiency. We compare our approach with flat and two-tiered models to demonstrate its viability. Central to our design is a node lookup and message routing mechanism, coupled with a consensus-accelerating protocol. Transaction, block verification, and creation processes involve a logarithmically bounded number of nodes per tier, improving consensus time and scalability as tier levels increase. The design concept maintains a balance between consensus decentralization and node overhead capacity, ensuring scalability and security. This design concept can be implemented using various algorithmic approaches to produce multiple proposed architectures, all leveraging the consensus-accelerating protocol. Simulations performed on one of the architectures provide solid proof of concept, demonstrating improved performance over flat and two-tier models. The proposed mechanisms provide general-purpose scalability, load balancing, node lookup, and message routing solutions, and are application- and consensus-algorithm-agnostic. © 2024 IEEE.

关键词： Efficiency

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：