检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Xu, Chengpei Jia, Wenjing Wang, Ruomei Luo, Xiaonan He, Xiangjian Faculty of Engineering and IT University of Technology Sydney Australia The Department of Computer Science University of Nottingham Ningbo China School of Data and Computer Science National Engineering Research Center of Digital Life Sun Yat-sen University China The School of Computer Science and Information Security Guilin University of Electronic Technology China

Bottom-up text detection methods play an important role in arbitrary-shape scene text detection but there are two restrictions preventing them from achieving their great potential, i.e., 1) the accumulation of false text segment detections, which affects subsequent processing, and 2) the difficulty of building reliable connections between text segments. Targeting these two problems, we propose a novel approach, named "MorphText", to capture the regularity of texts by embedding deep morphology for arbitrary-shape text detection. Towards this end, two deep morphological modules are designed to regularize text segments and determine the linkage between them. First, a Deep Morphological Opening (DMOP) module is constructed to remove false text segment detections generated in the feature extraction process. Then, a Deep Morphological Closing (DMCL) module is proposed to allow text instances of various shapes to stretch their morphology along their most significant orientation while deriving their connections. Extensive experiments conducted on four challenging benchmark datasets (CTW1500, Total-Text, MSRA-TD500 and ICDAR2017) demonstrate that our proposed MorphText outperforms both top-down and bottom-up state-of-the-art arbitrary-shape scene text detection approaches. Copyright © 2024, The Authors. All rights reserved.

关键词： Morphology

来源：评论

学校读者我要写书评

暂无评论

A New Federated Learning Framework Against Gradient Inversion Attacks 39

A New Federated Learning Framework Against Gradient Inversio...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Guo, Pengxin Zeng, Shuang Chen, Wenhao Zhang, Xiaodan Ren, Weihong Zhou, Yuyin Qu, Liangqiong School of Computing and Data Science The University of Hong Kong Hong Kong Department of Mathematics The University of Hong Kong Hong Kong College of Computer Science Beijing University of Technology China School of Mechanical Engineering and Automation Harbin Institute of Technology Shenzhen China Department of Computer Science and Engineering UC Santa Cruz United States

ISBN: (纸本)157735897X

Federated Learning (FL) aims to protect data privacy by enabling clients to collectively train machine learning models without sharing their raw data. However, recent studies demonstrate that information exchanged during FL is subject to Gradient Inversion Attacks (GIA) and, consequently, a variety of privacy-preserving methods have been integrated into FL to thwart such attacks, such as Secure Multi-party Computing (SMC), Homomorphic Encryption (HE), and Differential Privacy (DP). Despite their ability to protect data privacy, these approaches inherently involve substantial privacy-utility trade-offs. By revisiting the key to privacy exposure in FL under GIA, which lies in the frequent sharing of model gradients that contain private data, we take a new perspective by designing a novel privacy preserve FL framework that effectively "breaks the direct connection" between the shared parameters and the local private data to defend against GIA. Specifically, we propose a Hypernetwork Federated Learning (HyperFL) framework that utilizes hypernetworks to generate the parameters of the local model and only the hypernetwork parameters are uploaded to the server for aggregation. Theoretical analyses demonstrate the convergence rate of the proposed HyperFL, while extensive experimental results show the privacy-preserving capability and comparable performance of HyperFL. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Adaptive prescribed performance based on recursive nonsingular terminal sliding mode control for quad-rotor systems under uncertainty and disturbance: Real-time validation

引用

Aerospace science and Technology 2024年 147卷

作者： Han, Gao Mofid, Omid Mobayen, Saleh Khooban, Mohammad Hassan School of Aerospace Science and Technology Xidian University Xi'an710071 China Department of Electrical Engineering University of Zanjan Zanjan Iran Graduate School of Intelligent Data Science National Yunlin University of Science and Technology Yunlin Douliou Taiwan Department of Electrical and Computer Engineering Aarhus University Aarhus8200 Denmark

In this paper, at the aim of the fast trajectory-following control of the Unmanned Aerial Vehicle (UAV) systems subject to uncertainty and disturbance, the adaptive prescribed performance control based recursive nonsingular terminal sliding mode control is suggested. Afterward, for the fast trajectory-following control of the uncertain and perturbed quad-rotor system, a recursive nonsingular terminal sliding mode strategy is recommended. Since the better transient and steady-state response of the sliding mode surface is of the utmost importance in the design of the controller, the prescribed performance control scheme is proposed in which the transformed prescribed form of the previously proposed recursive nonsingular terminal sliding surface is obtained to prove that the recursive nonsingular terminal sliding surface can converge to a preset region. Whereas unknown bounded uncertainty and disturbance always exist in the UAV system's dynamical model which decreases its performance and efficiency, the adaptive mechanism is applied to approximate uncertain bounds of uncertainty and disturbance to advance the implementation of the UAV system. Later, quick reachability of the prescribed form switching surface is demonstrated using Lyapunov concept. Ultimately, simulation results using MATLAB/Simulink over a quad-rotor system and real-time simulation using the Speedgoat Real-Time Target Machine platform are conducted to validate the feasibility and performance of the proposed method. © 2024 Elsevier Masson SAS

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Explainable Artificial Intelligence in Malignant Lymphoma Classification: Optimized DenseNet121 Deep Learning Approach with Particle Swarm Optimization and Genetic Algorithm.

引用

IEEE Access 2025年 13卷 98639-98655页

作者： Elwahsh, Haitham Bakhiet, Ali Alirr, Omar Ibrahim Khalifa, Tarek Alsabaan, Maazen Ibrahem, Mohamed I. El-Shafeiy, Engy Applied Science Private University Faculty of Information Technology Amman 11931 Jordan Kafrelsheikh University Faculty of Computers and Information Department of Computer Science Kafrelsheikh 33516 Egypt Badya University Faculty of Computing and Data Science Egypt American University of the Middle East College of Engineering and Technology Egaila 54200 Kuwait King Saud University College of Computer and Information Sciences Department of Computer Engineering P.O.Box 51178 Riyadh 11543 Saudi Arabia Augusta University School of Computer and Cyber Sciences Augusta 30912 GA United States University of Sadat City Faculty of Computers and Artificial Intelligence Department of Computer Science Sadat City 32897 Egypt

One of the forms of cancerous tumors that can be fatal is malignant lymphoma. Histopathological examination of lymphoma tissue images is a diagnostic technique for detecting malignant lymphomas. Differentiating lymphoma subtypes manually is challenging due to their similar morphological features. This paper introduces a cutting-edge methodology for classifying malignant lymphoma using Hematoxylin and Eosin (H&E)-stained biopsy images. It integrates feature reduction using Particle Swarm Optimization (PSO) and model optimization through a Genetic Algorithm (GA) in conjunction with DenseNet121 architecture. The DenseNet121 model, which employs four distinct freezing strategies, extracts features from the images, generating an initial 4096 features per image. These features are subsequently reduced by more than 54.4% using PSO, thereby enhancing the computational efficiency without sacrificing classification accuracy. The trained GA Deep Neural Network (DNN) model achieved a significant score of 96.77% testing accuracy with high precision, recall and F1 scores as well. The results after comparison with other existing state-of-the-art models showed the usefulness of the proposed approach in terms of performance as well as the consumption of resources for medical imaging classification. It is recommended that the combined application of PSO for feature reduction and GA for model optimization can be successfully used for improving accuracy rate of such algorithms while reducing computation time. This technique is expected to be further developed and applied in different tasks of computer aided diagnosis of malignant lymphoma in particular the tasks with requirements for explainable artificial intelligence. © 2013 IEEE.

关键词： Artificial intelligent Deep Learning Genetic Algorithm malignant lymphoma Optimized DenseNet121 Particle Swarm Optimization

来源：评论

学校读者我要写书评

暂无评论

Forensic Investigation of Malicious Activities in Digital Environments

Forensic Investigation of Malicious Activities in Digital En...

引用

data Engineering and Communication Systems (ICDECS), International Conference on

作者： A. Pasumpon Pandian A.S. Anakath R. Kannadasan K. Ravikumar D Abdul Kareem Department of Computer Science and Engineering CARE College of Engineering Trichy India Department of Computer Science and Engineering Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Chennai Tamil Nadu India Vellore Tamil Nadu India Department of Computer Science & Engineering RRASE College of Engineering Padappai Chennai India Department of Artificial Intelligence and Data Science GRT Institute of Engineering and Technology Tiruttani Tamil Nadu India

Several digital dangers were investigated. Malware dominated analysis with 45 attacks. We found 30 phishing attacks. 22 data breaches, 15 cyber espionage, 18 identity theft. This indicates the kind and frequency of harassment digital forensic specialists endure. Digital evidence was crucial to the case. 65 pieces of evidence were collected on computers. 42 incidents showed smartphones and tablets’ digital investigative potential. There were 55 examples of communication patterns requiring network logs. In 28 and 12, cloud services and removable media provided proof. A wide range of evidence sources stresses the need for extensive collecting. Digital forensics were essential to the inquiry. In 70% of cases, disk imaging software created forensic copies of metadata-and timestamp-containing storage media. Network packet analyzers examined 48 traffic logs and communication patterns. To analyze volatile memory, 35 cases used memory forensics. In 25 cases, malware analysis platforms decrypted malware. In 60 situations, file recovery software recovered lost or hidden files. Digital forensic specialists’ diverse arsenal for thorough investigations is shown here. Legal and ethical considerations dominated the study. The law was important, and all inquiries followed norms. A strong chain of custody protects evidence integrity and admissibility. The data privacy rules were respected. Legal evidence is admissible in court. Ethics ensured objective investigations. This underlines digital forensic investigations’ legal and ethical foundation.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Performance Evaluation of Load Balancer in Storage Level Service Offering Applications 5th

Performance Evaluation of Load Balancer in Storage Level Ser...

引用

5th International Conference on Recent Developments in Control, Automation and Power Engineering, RDCAPE 2023

作者： Vianny, Maria Manuel Bhardwaj, Rajat Bhardwaj, Vivek Sharma, Vikrant Virender Manikandan, S. Department of Artificial Intelligence and Data Science Panimalar Engineering College Tamilnadu Chennai India Department of Computer Science and Engineering ASET Amity University Karnataka Bengaluru India School of Computer Science and Engineering Manipal University Jaipur Jaipur India Department of Computer Science and Engineering Graphic Era Hill University Dehradun India Chandigarh University Punjab Mohali India Department of Information Technology E.G.S. Pillay Engineering College Tamil Nadu Nagapattinam India

ISBN: (纸本)9789819600465

Cloud computing is a world view appropriately characterized for IT contributions where data and resources are aggregated by web-based cloud administration organizations using planned hardware and packages, primarily based on the web. People could use at a reasonable cost. Combining resources may affect the ease with which those resources are accessed, which might result in an issue. Load control is the process of distributing network traffic among several servers. No server will be overburdened as a result of this. In order to increase user responsiveness, load regulation equitably distributes the tasks. Additionally, it improves client accessibility to projects and websites. This archive is being kept in order to better understand charge control. It is connected to other structures via correspondence bodies like the Internet. Load balancing is an important distributed computing space to stay away from work beyond the workload and provide equally important support. Different calculations are used to determine the complexity of the system. Our suggested technique selects 100 virtual computers and, with 92% accuracy, determines where an ideal storage facility is for use in real time. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Cloud storage

来源：评论

学校读者我要写书评

暂无评论

Optimizing Synergy in Deep Learning for Accurate Diabetic Retinopathy Identification and Staging

Optimizing Synergy in Deep Learning for Accurate Diabetic Re...

引用

Electronics and Sustainable Communication Systems (ICESC), 2020 International Conference on

作者： Raghunadha Reddi Dornala Adusumilli Ramana Lakshmi Sudhir Ponnapalli Ganji Ramanjaiah Kopparthi Poorna Sairamakrishna Department Of Computer Science & Engineering PVP Siddhartha Engineering College India Dept.of Computer Science & Engineering (Data Science) R V R & J C College of Engineering Guntur Andhra Pradesh India Sagi Rama Krishna Raju Engineering College Bhimavaram

ISBN: (数字)9798350379945

ISBN: (纸本)9798350379952

Millions of people with diabetes suffer from diabetic retinopathy (DR), the primary cause of vision loss in human beings. Early detection and precise staging are essential to manage and treat DR effectively. In recent years, deep learning (DL) approaches have demonstrated encouraging results in automating the staging and detection of DR from retinal images. Developers have designed many existing systems to detect and classify DR images based on their staging. The proposed approach, also called a novel method, increases the detection rate regarding affected regions. Various architectures such as Residual Networks (ResNets), Attention Mechanisms, and Convolutional Neural Networks (CNNs) improved the performance of the proposed approach by obtaining significant data from retinal images. Transfer learning mainly transfers the disease and standard patterns from the training to testing models. An additional advantage of the proposed approach is that it combines advanced preprocessing and feature extraction techniques. The preprocessing technique reduces the noise and filters the images without losing the pixel quality. The feature extraction helps extract the essential patterns from the input DR images that help detect the affected regions. Experimental analyses were applied using two publicly accessible datasets, such as the Kaggle DR dataset and Kaggle Fundus images, which contain various retinal pictures with different levels of DR severity. Comparing our results against state-of-the-art methods, we find considerable identification and stage accuracy increases.

关键词： Deep learning Diabetic retinopathy Accuracy Attention mechanisms Noise Transfer learning Feature extraction Retina Convolutional neural networks Diseases

来源：评论

学校读者我要写书评

暂无评论

FedMUA: Exploring the Vulnerabilities of Federated Learning to Malicious Unlearning Attacks

arXiv

引用

arXiv 2025年

作者： Chen, Jian Lin, Zehui Lin, Wanyu Shi, Wenlong Yin, Xiaoyan Wang, Di Department of Data Science and Artificial Intelligence The Hong Kong Polytechnic University Hong Kong Department of Data Science and Artificial Intelligence The Department of Computing The Hong Kong Polytechnic University Hong Kong School of Electronic Information and Communications Huazhong University of Science and Technology Wuhan430074 China School of Information Science and Technology Northwest University Xi’an China Division of Computer Electrical and Mathematical Sciences and Engineering King Abdullah University of Science and Technology Saudi Arabia

Recently, the practical needs of "the right to be forgotten" in federated learning gave birth to a paradigm known as federated unlearning, which enables the server to forget personal data upon the client’s removal request. Existing studies on federated unlearning have primarily focused on efficiently eliminating the influence of requested data from the client’s model without retraining from scratch, however, they have rarely doubted the reliability of the global model posed by the discrepancy between its prediction performance before and after unlearning. To bridge this gap, we take the first step by introducing a novel malicious unlearning attack dubbed FedMUA, aiming to unveil potential vulnerabilities emerging from federated learning during the unlearning process. Specifically, clients may act as attackers by crafting malicious unlearning requests to manipulate the prediction behavior of the global model. The crux of FedMUA is to mislead the global model into unlearning more information associated with the influential samples for the target sample than anticipated, thus inducing adverse effects on target samples from other clients. To achieve this, we design a novel two-step method, known as Influential Sample Identification and Malicious Unlearning Generation, to identify and subsequently generate malicious feature unlearning requests within the influential samples. By doing so, we can significantly alter the predictions pertaining to the target sample by initiating the malicious feature unlearning requests, leading to the deliberate manipulation for the user adversely. Additionally, we design a new defense mechanism that is highly resilient against malicious unlearning attacks. Extensive experiments on three realistic datasets reveal that FedMUA effectively induces misclassification on target samples and can achieve an 80% attack success rate by triggering only 0.3% malicious unlearning requests. Copyright © 2025, The Authors. All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Leveraging Machine Learning and Forecasting Techniques to Enhance Credit Risk Analysis and Prediction

Leveraging Machine Learning and Forecasting Techniques to En...

引用

Self Sustainable Artificial Intelligence Systems (ICSSAS), International Conference on

作者： S Satheeshkumar M Dakshana K Gunalan P Anandan R Saveetha M Nithya Department of Artificial intelligence and data science Bannari Amman Institute of Technology Tamilnadu India Department of Physics Sathyabama Institute of Science and Technology Tamilnadu India Department of Computer Technology Bannari Amman Institute of Technology Tamilnadu India

ISBN: (数字)9798350368413

ISBN: (纸本)9798350368420

This research presents a novel data-driven framework to improve credit risk assessment and minimize loan defaults. The framework leverages a combination of machine learning techniques, including Random Forest and XGBoost, to analyze a comprehensive dataset from Axis Bank spanning 2007-2015. The study focuses on key factors such as loan amount, debt-to-income ratio, interest rate, and debt burden to predict the likelihood of default. By employing oversampling techniques to address data imbalance and integrating ensemble methods, the framework enhances prediction accuracy. The results demonstrate the superior performance of the XGBoost model in accurately classifying loans as default or non-default. This research contributes to the advancement of credit risk management by providing a powerful tool for financial institutions to make informed lending decisions and mitigate financial risks.

关键词： Accuracy Learning (artificial intelligence) Predictive models Market research Ensemble learning Forecasting Random forests Monitoring Long short term memory Risk mitigation

来源：评论

学校读者我要写书评

暂无评论

LEARNING LONG-TERM MUSIC REPRESENTATIONS VIA HIERARCHICAL CONTEXTUAL CONSTRAINTS 22

LEARNING LONG-TERM MUSIC REPRESENTATIONS VIA HIERARCHICAL CO...

引用

22nd International Conference on Music Information Retrieval, ISMIR 2021

作者： Wei, Shiqi Xia, Gus School of Data Science Fudan University China Music X Lab Computer Science Department New York University Shanghai China

ISBN: (纸本)9781732729902

Learning symbolic music representations, especially disentangled representations with probabilistic interpretations, has been shown to benefit both music understanding and generation. However, most models are only applicable to short-term music, while learning long-term music representations remains a challenging task. We have seen several studies attempting to learn hierarchical representations directly in an end-to-end manner, but these models have not been able to achieve the desired results and the training process is not stable. In this paper, we propose a novel approach to learn long-term symbolic music representations through contextual constraints. First, we use contrastive learning to pre-train a long-term representation by constraining its difference from the short-term representation (extracted by an off-the-shelf model). Then, we fine-tune the long-term representation by a hierarchical prediction model such that a good long-term representation (e.g., an 8-bar representation) can reconstruct the corresponding short-term ones (e.g., the 2-bar representations within the 8-bar range). Experiments show that our method stabilizes the training and the fine-tuning steps. In addition, the designed contextual constraints benefit both reconstruction and disentanglement, significantly outperforming the baselines. © 2021 Proceedings of the 22nd International Conference on Music Information Retrieval, ISMIR 2021. All Rights Reserved.

关键词： Music

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：