检索结果-内蒙古大学图书馆

Variable-Stepsize Multilayer Neural Network for Subpixel Target Detection in Hyperspectral Imaging

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 2025年 18卷 1718-1733页

作者： Lo, Edisanter Susquehanna Univ Dept Math & Comp Sci Selinsgrove PA 17870 USA

Conventional algorithms for subpixel target detection of a rare target in hyperspectral imaging are derived from the generalized likelihood ratio test. Artificial neural networks are designed for classification, which needs large samples of pixels for training background and target classes, but have a problem with subpixel target detection, which typically has only one target pixel for training the target class. The current detection algorithm for subpixel target detection based on neural networks uses a single-layer neural network and stochastic gradient descent method with variable stepsize to solve the optimization problem. The objective of this paper is to improve the current algorithm by developing a detection algorithm for subpixel target detection based on a multilayer neural network and stochastic gradient descent method with variable stepsize. The decision boundary is linear for the single-layer neural network and nonlinear for the multi-layer neural network. Experimental results from two hyperspectral images, one with simulated subpixel target pixels for validating the algorithm and the other with actual subpixel target pixels for validation, have shown that the proposed algorithm can perform better than the conventional algorithms.

关键词： Vectors Neural networks Object detection Hyperspectral imaging Training Linear programming classification algorithms Multi-layer neural network Earth Training data neural network stochastic gradient descent (SGD) target detection

来源：评论

学校读者我要写书评

暂无评论

Double Oracle Neural Architecture Search for Game Theoretic Deep Learning Models

引用

IEEE TRANSACTIONS ON IMAGE PROCESSING 2025年 34卷 2463-2472页

作者： Aung, Aye Phyu Phyu Wang, Xinrun Wang, Ruiyu Chan, Hau An, Bo Li, Xiaoli Senthilnath, J. ASTAR Inst Infocomm Res Singapore 138632 Singapore Singapore Management Univ Sch Comp & Informat Syst Singapore 188065 Singapore KTH Royal Inst Technol Div Robot Percept & Learning S-10044 Stockholm Sweden Univ Nebraska Lincoln Sch Comp Lincoln NE 68588 USA Nanyang Technol Univ Sch Comp Sci & Engn Singapore 639798 Singapore ASTAR Ctr Frontier AI Res Singapore 138632 Singapore

In this paper, we propose a new approach to train deep learning models using game theory concepts including Generative Adversarial Networks (GANs) and Adversarial Training (AT) where we deploy a double-oracle framework using best response oracles. GAN is essentially a two-player zero-sum game between the generator and the discriminator. The same concept can be applied to AT with attacker and classifier as players. Training these models is challenging as a pure Nash equilibrium may not exist and even finding the mixed Nash equilibrium is difficult as training algorithms for both GAN and AT have a large-scale strategy space. Extending our preliminary model DO-GAN, we propose the methods to apply the double oracle framework concept to Adversarial Neural Architecture Search (NAS for GAN) and Adversarial Training (NAS for AT) algorithms. We first generalize the players' strategies as the trained models of generator and discriminator from the best response oracles. We then compute the meta-strategies using a linear program. For scalability of the framework where multiple network models of best responses are stored in the memory, we prune the weakly-dominated players' strategies to keep the oracles from becoming intractable. Finally, we conduct experiments on MNIST, CIFAR-10 and TinyImageNet for DONAS-GAN. We also evaluate the robustness under FGSM and PGD attacks on CIFAR-10, SVHN and TinyImageNet for DONAS-AT. We show that all our variants have significant improvements in both subjective qualitative evaluation and quantitative metrics, compared with their respective base architectures.

关键词： Games Generative adversarial networks Generators Training Computer architecture Neural architecture search Robustness Deep learning classification algorithms Terminology neural architecture search adversarial training double oracle game theory

来源：评论

学校读者我要写书评

暂无评论

An Opposition-Based Great Wall Construction Metaheuristic Algorithm With Gaussian Mutation for Feature Selection

引用

IEEE ACCESS 2024年 12卷 30796-30823页

作者： Zitouni, Farouq Almazyad, Abdulaziz S. Xiong, Guojiang Mohamed, Ali Wagdy Harous, Saad Kasdi Merbah Univ Dept Comp Sci & Informat Technol Ouargla 30000 Algeria King Saud Univ Coll Comp & Informat Sci Dept Comp Engn Riyadh 11543 Saudi Arabia Guizhou Univ Coll Elect Engn Guizhou Key Lab Intelligent Technol Power Syst Guiyang 550025 Peoples R China Cairo Univ Fac Grad Studies Stat Res Operat Res Dept Giza 12613 Egypt Appl Sci Private Univ Appl Sci Res Ctr Amman 11931 Jordan Univ Sharjah Coll Comp & Informat Dept Comp Sci Sharjah U Arab Emirates

The feature selection problem involves selecting a subset of relevant features to enhance the performance of machine learning models, crucial for achieving model accuracy. Its complexity arises from the vast search space, necessitating the application of metaheuristic methods to efficiently identify optimal feature subsets. In this work, we employed a recently proposed metaheuristic algorithm named the Great Wall Construction Algorithm to address this challenge - a powerful optimizer with promising results. To enhance the algorithm's performance in terms of exploration, exploitation, and avoidance of local optima, we integrated opposition-based learning and Gaussian mutation techniques. The proposed algorithm underwent a comprehensive comparative analysis against ten influential state-of-the-art methodologies, encompassing seven contemporary algorithms and three classical counterparts. The evaluation covered 22 datasets of varying sizes, ranging from 9 to 856 features, and included the utilization of six distinct evaluation metrics related to accuracy, classification error rate, number of selected features, and completion time to facilitate comprehensive comparisons. The obtained numerical results underwent rigorous scrutiny through several non-parametric statistical tests, including the Friedman test, the post hoc Dunn's test, and the Wilcoxon signed ranks test. The resulting mean ranks and p-values unequivocally demonstrate the superior efficacy of the proposed algorithm in addressing the feature selection problem. The Matlab source code for the proposed approach is available for access via the link "https://***/matlabcentral/fileexchange/159728-an-opposition-based-gwca-for-thefs-problem".

关键词： classification algorithms Metaheuristics Transfer functions Feature extraction Computational modeling Convergence Gaussian processes Source coding Machine learning Matlab Feature selection problem great wall construction metaheuristic algorithm opposition-based learning Gaussian mutation

来源：评论

学校读者我要写书评

暂无评论

A Nonstationary Clutter Suppression Algorithm for Space-Based Radar Based on Presequence Pulse Echoes

引用

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS 2025年第1期61卷 542-553页

作者： Wu, Jiaye Chen, Wei Liu, Weijian Zhang, Tianfu Zhang, Zhaojian Wang, Yong-Liang Wuhan Univ Sch Elect Informat Wuhan 430072 Peoples R China Wuhan Univ Elect Informat Sch Wuhan 430072 Peoples R China Wuhan Elect Informat Inst Wuhan 430072 Peoples R China Wuhan Elect Informat Inst Wuhan 430019 Peoples R China Chinese Acad Sci Aerosp Informat Res Inst Beijing 100094 Peoples R China

Owing to the Earth's rotation and the rapid motion of satellites, the clutter received by space-based radar (SBR) exhibits severe nonstationary characteristics, notably evidenced by the Doppler frequency of clutter varying with range, particularly in the near-range region. Elevation prefiltering is shown to be an effective algorithm for suppressing near-range nonstationary clutter. Based on the characteristics of SBR with several near-range bending clutters, this article proposes an elevation prefiltering algorithm based on presequence pulse echoes to suppress nonstationary clutter for SBR. The training samples selected only contain near-range nonstationary clutter, which is obtained from the elevation dimension data of the presequence pulse echoes. By excluding the influence of far-range clutter, the proposed algorithm demonstrates superior performance in suppressing near-range clutter. After attenuating the near-range nonstationary clutter, the azimuth-Doppler space-time adaptive processing algorithm is cascaded to further suppress far-range stationary clutter. Simulation results show the effectiveness in suppressing near-range clutter and enhancing target detection capabilities for SBR of the proposed method.

关键词： Clutter Training Spaceborne radar Doppler effect Radar clutter Logic gates classification algorithms

来源：评论

学校读者我要写书评

暂无评论

Implementation and Performance Evaluation of Machine Learning-Based Apriori Algorithm to Detect Non-Technical Losses in Distribution Systems

引用

IEEE ACCESS 2025年 13卷 32289-32305页

作者： Akhtar, Muhammad Faheem Farooq, Haroon Akhtar, Muhammad Naveed Hussain, Ghulam Amjad Kashif, Syed Abdul Rahman Rashid, Zeeshan Safdar, Madia Univ Engn & Technol Dept Elect Engn Lahore 39161 Pakistan Rachna Coll Engn & Technol Dept Elect Engn Gujranwala 52250 Pakistan Univ Dubai Coll Engn & IT Dubai U Arab Emirates Islamia Univ Bahawalpur Dept Elect Engn Bahawalpur 63100 Pakistan Lappeenranta Lahti Univ Technol LUT Dept Elect Engn Lappeenranta 53850 Finland

The emergence and augmentation of nontechnical losses (NTLs) in a power distribution system has always been considered a critical issue in the global electricity market. Being considered as uncharged, unlawfully or unfairly charged consumed electricity, NTLs due to fraudulent activities can impose a serious burden on the grid and revenue loss in the state's budget. The detection of NTLs based on the analysis of huge consumer dataset using machine learning is an acknowledged approach due to its expedient performance compared to manual inspection. This paper addresses a comprehensive investigation of monthly electricity consumption data of similar to 15000 consumers over three years using machine learning for NTL detection. The data is acquired by meters having automatic meter reading (AMR) capability and is processed using support vector machine (SVM) classifier, deep neural network (DNN), gradient boosted reinforcement learning (GBRL) and apriori algorithm. The results of ML algorithms are assessed by various performance metrics based on the confusion matrix and compared among themselves as well as with the findings from the published works. The least recall score of 0.8926 is exhibited by SVM classifier and poorest scores of accuracy (0.9376), specificity (0.9451), precision (0.7521), F1 (0.8183) and false positive rate (0.0549) are given by DNN program. With apriori algorithm, the scores of accuracy, recall, specificity, precision, F1 and false positive rate are observed as 0.9964, 0.9927, 0.9971, 0.9843, 0.9885, 0.0029 respectively. In all the performance based six domains, a significant improvement of 6%, 10%, 5%, 23%, 17% and 5% respectively compared to the least scores is demonstrated by the apriori algorithm. Therefore, in all these domains, the apriori algorithm being reported for the first time in this research work outperforms all other methods of this paper as well as the similar published works.

关键词： Machine learning neural networks support vector machines classification algorithms unsupervised learning energy consumption.

来源：评论

学校读者我要写书评

暂无评论

Variable Action Set-Based Monte Carlo Tree Search Algorithm for AAV Autonomous Collision Avoidance

引用

IEEE ACCESS 2025年 13卷 43861-43877页

作者： Xu, Fulong Li, Bo Civil Aviat Flight Univ China Coll Air Traff Management Guanghan 618300 Peoples R China

Autonomous collision avoidance of Autonomous Aerial Vehicles (AAVs) is becoming a hot research topic in Urban Air Mobility (UAM) operations. Many studies modeled this problem as a Markov decision process (MDP) and adopted intelligent methods such as reinforcement learning and Monte Carlo Tree Search (MCTS) to design collision avoidance algorithms. Among these studies, some algorithms with discrete action sets have achieved good results in collision avoidance. However, the fixed discrete action set often leads to low flight quality, as the rigid control makes the AAV unable to stably fly straight in the desired direction but can only fly wobbly. To address this problem, a Variable Action Set-based MCTS algorithm (VASM) for autonomous collision avoidance is proposed in this paper. By introducing an additional action option that is carefully designed to improve the stability and efficiency of flight along any desired heading, a variable action set is constructed for VASM. In addition, a reward function that takes into account more rational evaluation factors is also proposed to facilitate better decision-making. We conducted numerical experiments to verify the effectiveness of VASM, and results show that compared with the conventional MCTS algorithm, VASM can better improve flight control and flight quality with only a small increase in computational cost, achieving more efficient and energy-saving flight guidance. Specifically, in high-density airspace, VASM reduces average flight distance by 15.34%, average flight time by 15.08%, average flight turns by 36.64%, conflict probability by 18.56%, while increasing target arrival probability to 98.3%.

关键词： Collision avoidance Autonomous aerial vehicles Aircraft Real-time systems Trajectory Monte Carlo methods Energy efficiency classification algorithms Atmospheric modeling Turning Autonomous collision avoidance Monte Carlo tree search Markov decision process AAV variable action set

来源：评论

学校读者我要写书评

暂无评论

Geographical Origin Identification of Panax notoginseng Using a Modified K-Nearest Neighbors Model With Near-Infrared Spectroscopy

引用

IEEE ACCESS 2025年 13卷 13832-13846页

作者： Cheng, Xuefeng Liao, Min Liu, Juan Chongqing City Management Coll Sch Big Data & Informat Ind Chongqing 401331 Peoples R China Chongqing Prod Council Chongqing 401147 Peoples R China Chongqing Univ Educ Sch Artificial Intelligence Chongqing 400065 Peoples R China Chongqing Univ Chongqing 400044 Peoples R China Chinese Acad Sci Chongqing Inst Green & Intelligent Technol Chongqing 400714 Peoples R China Univ Chinese Acad Sci Chongqing Coll Chongqing 400722 Peoples R China

This study introduces a novel method for identifying the geographical origins of Panax notoginseng using near-infrared spectroscopy and a modified K-nearest neighbors algorithm. The proposed model integrates distance and cosine similarity metrics to enhance classification performance, particularly for imbalanced datasets. Principal component analysis is employed to reduce dimensionality, significantly improving computational efficiency without sacrificing accuracy. Comparative analyses reveal that the modified K-nearest neighbors surpasses classic K-nearest neighbors and other models, achieving up to 96.90% accuracy on balanced datasets and 95.79% on imbalanced datasets. These results demonstrate the potential of combining spectral data processing with advanced machine learning techniques for efficient and accurate geographical origin identification, providing a robust tool for quality assurance and traceability in traditional Chinese medicine.

关键词： Nearest neighbor methods Accuracy Powders Analytical models Spectroscopy Data models classification algorithms Principal component analysis Chemicals Mass spectroscopy Panax notoginseng geographical origin identification near-infrared spectroscopy modified K-nearest neighbors algorithm

来源：评论

学校读者我要写书评

暂无评论

Cutting-Edge Intrusion Detection in IoT Networks: A Focus on Ensemble Models

引用

IEEE ACCESS 2025年 13卷 8375-8392页

作者： Sama, Najm Us Ullah, Saeed Kazmi, S. M. Ahsan Mazzara, Manuel Univ Derby Sch Comp Derby DE22 3AW England Univ West England Sch Comp Sci & Creat Technol Bristol BS16 1QY England Innopolis Univ Inst Software Dev & Engn Innopolis 420500 Russia

As the Internet of Things (IoT) landscape rapidly evolves, robust network security measures are imperative. In particular, Intrusion Detection Systems play a very important role in the preservation of an IoT environment from malicious activities. This paper provides a comprehensive performance comparison of various machine learning classifiers, including K-Nearest Neighbors, Gradient Boosting, XGBoost, Support Vector Machines, Random Forests, Decision Trees, and Extremely Randomized Trees, for intrusion detection in IoT networks. Comparative analysis shows that although all models did very well, the ensemble methods-GB, XGBoost, RF, and ERT-constantly performed better than others in F1-Score, recall, accuracy, and precision. Among them, ERT is turned out to be the most effective model for real-time attack detection on IoT devices, with an accuracy of 99.7% besides excellent precision and recall. XGBoost and RF also turn out to have high reliability and accuracy with F1-Scores of 0.95. These findings further underscore that ensemble methods outperform in intrusion detection for IoT networks and, thus, offer important insights to improve security within networks and protect critical IoT-based infrastructures from a variety of threats.

关键词： Internet of Things Accuracy Intrusion detection Machine learning Feature extraction Security Nearest neighbor methods Support vector machines Organizations classification algorithms Internet of Things (IoT) intrusion detection systems (IDS) machine learning classifiers

来源：评论

学校读者我要写书评

暂无评论

Integrating Advanced Techniques: RFE-SVM Feature Engineering and Nelder-Mead Optimized XGBoost for Accurate Lung Cancer Prediction

引用

IEEE ACCESS 2025年 13卷 29589-29600页

作者： Ayad, Sarah Al-Jamimi, Hamdi A. El Kheir, Ammar Arab Open Univ Fac Comp Studies Riyadh Saudi Arabia King Fahd Univ Petr Minerals Informat & Comp Sci Dhahran 31216 Saudi Arabia King Fahd Univ Petr Minerals Res Excellence Dhahran 31216 Saudi Arabia Liege Univ Hosp Ctr Dept Pediat Liege Belgium

Early detection of lung cancer is crucial for improving patient survival and reducing mortality. However, medical datasets often face challenges like irrelevant features and class imbalance, complicating accurate predictions. This study presents a comprehensive AI-powered lung cancer classification approach that enhances predictive accuracy and treatment planning. Our methodology combines Recursive Feature Elimination with Support Vector Machines (RFE-SVM) for effective feature selection and employs the XGBoost ensemble learning algorithm for classification, optimized using the Nelder-Mead algorithm. Evaluating the model's generalizability on two distinct lung cancer datasets, results show that our approach outperforms traditional machine learning models, achieving 100% accuracy. This research highlights the importance of advanced computational techniques in healthcare, paving the way for more personalized and effective patient care.

关键词： Lung cancer Cancer Accuracy Lungs Prediction algorithms Feature extraction Predictive models Medical diagnostic imaging classification algorithms Support vector machines Early prediction feature engineering lung cancer XGBoost

来源：评论

学校读者我要写书评

暂无评论

Synthetic Data Generation and Evaluation Techniques for Classifiers in Data Starved Medical Applications

引用

IEEE ACCESS 2025年 13卷 16584-16602页

作者： Bae, Wan D. Alkobaisi, Shayma Horak, Matthew Bankar, Siddheshwari Bhuvaji, Sartaj Kim, Sungroul Park, Choon-Sik Seattle Univ Dept Comp Sci Seattle WA 98122 USA United Arab Emirates Univ Coll Informat Technol Al Ain U Arab Emirates Amazon AWS Lambda Seattle WA USA Soonchunhyang Univ Grad Sch Dept ICT Environm Hlth Syst Asan 31538 South Korea Soonchunhyang Univ Bucheon Hosp Dept Internal Med Bucheon 14584 South Korea

With their ability to find solutions among complex relationships of variables, machine learning (ML) techniques are becoming more applicable to various fields, including health risk prediction. However, prediction models are sensitive to the size and distribution of the data they are trained on. ML algorithms rely heavily on vast quantities of training data to make accurate predictions. Ideally, the dataset should have an equal number of samples for each label to encourage the model to make predictions based on the input data rather than the distribution of the training data. In medical applications, class imbalance is a common issue because the occurrence of a disease or risk episode is often rare. This leads to a training dataset where healthy cases outnumber unhealthy ones, resulting in biased prediction models that struggle to detect the minority, unhealthy cases effectively. This paper addresses the problem of class imbalance, given the scarcity of training datasets by improving the quality of generated data. We propose an incremental synthetic data generation system that improves data quality over iterations by gradually adjusting to the data distribution and thus avoids overfitting in classifiers. Through extensive experimental assessments on real asthma patients' datasets, we demonstrate the efficiency and applicability of our proposed system for individual-based health risk prediction models. Incremental SMOTE methods were compared to the original SMOTE variants as well as various architectures of autoencoders. Our incremental data generation system enhances selected state-of-the-art SMOTE methods, resulting in sensitivity improvements for deep transfer learning (TL) classifiers ranging from 4.01% to 7.79%. Compared with the performance of TL without oversampling, the improvement achieved by the incremental SMOTE methods ranged from 27.18% to 40.97%. These results highlight the effectiveness of our technique in predicting asthma risk and their applicability

关键词： Autoencoders Synthetic data Predictive models Data models Medical services classification algorithms Asthma Training data Generative adversarial networks Boosting class imbalance problem control coefficient data starved contexts rare event prediction synthetic minority oversampling technique transfer learning

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：