检索结果-内蒙古大学图书馆

Dual-stream autoencoder for channel-level multi-scale feature extraction in hyperspectral unmixing

KNOWLEDGE-BASED SYSTEMS 2025年 317卷

作者： Gan, Yuquan Wang, Yong Li, Qiuyu Luo, Yiming Wang, Yihong Pan, Yushan Xian Univ Posts & Telecommun Sch Telecommun & Informat Engn Xian 710121 Shaanxi Peoples R China Xian Univ Posts & Telecommun Ctr Image & Informat Proc Xian 710121 Shaanxi Peoples R China Xian Jiaotong Liverpool Univ Dept Comp Cognit Intelligence & Creat Comp Grp Suzhou 215123 Jiangsu Peoples R China Xidian Univ Sch Cyber Engn Xian 710125 Peoples R China

High-dimensional hyperspectral imagery presents significant challenges for accurate unmixing due to spectral variability, limited spatial resolution, and noise. Traditional unmixing approaches often rely on spatial multi-scale processing, leading to redundant computations and suboptimal feature representations. In response to these challenges, we propose a novel Channel Multi-Scale Dual-Stream autoencoder (CMSDAE) that innovatively integrates channel-level multi-scale feature extraction with dedicated spectral information guidance. By leveraging Channel-level Multi-Scale Perception Blocks and a Hybrid Attention-Aware Feature Block, CMSDAE efficiently captures diverse and robust spectral-spatial features while significantly reducing computational redundancy. Extensive experiments on both synthetic and real-world datasets demonstrate that CMSDAE not only improves unmixing accuracy and robustness against noise but also offers enhanced computational efficiency compared to state-of-the-art methods. This work provides new insights into spectral-spatial modeling for hyperspectral unmixing, promising more reliable and scalable analysis in challenging remote sensing applications.

关键词： Hyperspectral unmixing Multi-scale feature extraction Channel-level multi-scale Spectral information guidance autoencoder

来源：评论

学校读者我要写书评

暂无评论

Landmark Block-Embedded Aggregation autoencoder for Anomaly Detection

引用

IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS 2025年第2期55卷 1004-1019页

作者： Liu, Ye Tian, Yuanrong Mi, Yunlong Liu, Hui Wang, Jianqiang Pedrycz, Witold Cent South Univ Sch Business Changsha 410083 Peoples R China Natl Univ Def Technol Sch Elect Countermeasure Hefei 230037 Peoples R China Univ Alberta Dept Elect & Comp Engn Edmonton AB T6G 2R3 Canada Polish Acad Sci Syst Res Inst PL-00901 Warsaw Poland Istinye Univ Fac Engn & Nat Sci Dept Comp Engn TR-34010 Istanbul Turkiye

Unsupervised anomaly detection (AD) methods based on deep learning have attracted great attention in unlabeled data mining. The performance of these AD methods usually depends on the representation ability of normal patterns and the quality of training data. However, most deep unsupervised AD methods do not capture the distribution characteristics and the diversity of normal patterns effectively. In the meantime, they ignore the interference of abnormal samples on the model in training data with anomaly contamination. To tackle these issues, this article proposes a method named landmark block-embedded aggregation autoencoder (LBAA) for AD. LBAA constructs a filter and an aggregation autoencoder by introducing a novel normal feature learning approach to improve data quality and adjust its distribution differences from anomalies. In the normal feature learning, we define a landmark block to represent distribution of a normal class and an adaptive selection mechanism of landmark blocks' number to obtain diverse normal features. On the basis, the filter is constructed to filter distinct anomalies and improve the quality of the contaminated training data. Then, a weighted objective function is proposed to train the aggregation autoencoder. The function can reduce the interference of anomalies and realize the aggregation of normal samples to increase the feature differences between normal and abnormal samples. Next, the trained aggregation autoencoder calculates the anomaly score of each sample by summing the reconstruction error and its median sparseness to the landmark blocks. Finally, we report on a comprehensive experiment on multiple datasets. The obtained results validate the effectiveness and robustness of LBAA.

关键词： Anomaly detection (AD) autoencoder landmark block landmark block unsupervised learning unsupervised learning unsupervised learning

来源：评论

学校读者我要写书评

暂无评论

Multimodal geometric autoencoder (MGAE) for rail fasteners tightness evaluation with point clouds & monocular depth fusion

引用

MEASUREMENT 2025年 244卷

作者： Qiu, Shi Zaheer, Qasim Shah, S. Muhammad Ahmed Hassan Shah, Syed Faizan Hussain Ehsan, Haleema Atta, Zunaira Ai, Chengbo Wang, Jin Wang, Weidong Peng, Jun Cent South Univ Sch Civil Engn Changsha 410075 Peoples R China Cent South Univ MOE Key Lab Engn Struct Heavy Haul Railway Changsha 410075 Peoples R China Cent South Univ Ctr Railway Infrastruct Smart Monitoring & Managem Changsha 410075 Peoples R China Univ Massachusetts Dept Civil & Environm Engn Amherst MA 01003 USA

Accurate detection and estimation of railway fastener tightness are vital for rail infrastructure safety and reliability. Traditional methods depend on manual annotation tools like Label Me, which are error-prone, laborintensive, and costly. Additionally, monocular depth estimation and instance segmentation involve complex computations that challenge real-time implementation, particularly on resource-constrained platforms. This study introduces a novel three-phase solution using the Multimodal Geometric autoencoder (MGAE) for fastener tightness detection, integrating point clouds with monocular-depth-guided multimodal data. Our approach utilizes a hybrid autoencoder for high-quality feature extraction, enabling precise tightness estimation. Employing unsupervised learning, MGAE eliminates the need for labeled data, thus reducing labor and costs. The framework integrates point clouds, mesh, monocular depth, and 2D images, with various fusion blocks enhancing feature extraction accuracy and computational efficiency. Post-feature extraction, classical techniques such as isolation forest, stress-strain, and elastic potential energy methods assess fastener tightness.

关键词： Rail fasteners Fastener tightness Rail infrastructure autoencoder Multimodal approach

来源：评论

学校读者我要写书评

暂无评论

Attack stage detection method based on vector reconstruction error autoencoder and explainable artificial intelligence

引用

JOURNAL OF SUPERCOMPUTING 2025年第1期81卷 1-37页

作者： Li, Jinze Meng, Xiangyu Qi, Zichen Guo, Dong Fu, Cong Jilin Univ Coll Software Changchun 130012 Peoples R China Jilin Univ Coll Comp Sci & Technol Changchun 130012 Peoples R China Jilin Univ Minist Educ Symbol Computat & Knowledge Changchun 130012 Peoples R China

One of the most serious security threats faced by the Internet today is multi-stage attacks. In response to this challenge, anomaly detection-based methods have been widely used to identify different stages of such attacks. However, current anomaly detection approaches for detecting attack stages face several challenges: (1) Traditional methods often adopt a global perspective, lacking detailed consideration of the traffic characteristics at each stage, which may reduce the accuracy in detecting specific stages. (2) Many detection methods rely on deep learning models with complex internal structures, making their decision-making process opaque and difficult for users to interpret. This also complicates model optimization and improvement. To address these challenges, this paper proposes an attack stage detection method based on a vector reconstruction error autoencoder. By analyzing each stage independently, the proposed method enhances detection precision. It also integrates the permutation feature importance technique to quantify and interpret the model's reliance on different features, guiding feature selection and model optimization. Experiments conducted using the CIC-IDS2017 and CSE-CIC-IDS2018 datasets demonstrate that the proposed method achieves higher accuracy, precision, recall, and F1 score compared to other methods, confirming its feasibility and effectiveness in detecting multi-stage attacks.

关键词： Multi-stage attack Attack stage detection autoencoder Explainable artificial intelligence Intrusion detection

来源：评论

学校读者我要写书评

暂无评论

An autoencoder-based recommendation framework toward cold start problem

引用

JOURNAL OF SUPERCOMPUTING 2025年第1期81卷 1-24页

作者： Zhou, Wang Tian, Ying Haq, Amin Ul Ahmad, Sultan Xihua Univ Sch Comp & Software Engn Chengdu Peoples R China Sichuan Police Coll Intelligent Policing Key Lab Luzhou Peoples R China Univ Elect Sci & Technol China Sch Comp Sci & Engn Chengdu Peoples R China Prince Sattam Bin Abdulaziz Univ Coll Comp Engn & Sci Dept Comp Sci Alkharj Saudi Arabia

Recommender methods have been effectively used in both academic and industrial settings. However, the cold start problem with scarce prior information has become the barrier hindering recommender systems from gaining further improvements. To overcome this issue, this article proposes a novel autoencoder framework referred to as CSRec, which owns the merits of both neural networks and collaborative filtering. Specifically, to search the nearest neighbors for cold start items, CSRec learns item representation and carries out clustering for items via k-means++ method. After that, with the nearest item cluster, CSRec could perform rating prediction for the cold start items through the autoencoder architecture, which could reconstruct the input space directly. Identically, CSRec could also perform cold start recommendations for users through the neural network. In practice, with the autoencoder architecture, CSRec owns powerful capability in computation and representation, which could deeply exploit the inner relationship for items and yield high performance in addressing cold start issues. Moreover, it could enhance the novelty and diversity of cold start recommendations. Experiments on CiaoDVDs and DoubanMovie certificate the superiority of CSRec in addressing the cold start issue, which could yield accurate performance in terms of RMSE, MAE and top-K and outperform other benchmark recommender approaches significantly.

关键词： Neural network Cold start autoencoder Collaborative filtering

来源：评论

学校读者我要写书评

暂无评论

Stacked dynamic target regularization enhanced autoencoder for soft sensor in industrial processes

引用

CANADIAN JOURNAL OF CHEMICAL ENGINEERING 2025年第3期103卷 1335-1348页

作者： Guo, Xiaoping Zhao, Xiaofeng Li, Yuan Shenyang Univ Chem Technol Coll Informat Engn Shenyang Peoples R China

Stacked autoencoders (SAEs) have great potential in developing soft sensors due to their excellent feature extraction capabilities. However, the pre-training stage of SAE is unsupervised and some important information related to target variables may be discarded. Meanwhile, as the depth of the network increases, reconstruction errors continue to accumulate, resulting in incomplete feature representations of the original input. In addition, the dynamic nature of the data affects the predictive results of the model. To address these issues, the stacked dynamic target regularization enhanced autoencoder (SDTR-EAE) method is proposed, which adds the DTR and the original input information layer by layer to enhance the feature extraction. To adapt to the dynamic changes in data and extract target-related features, entropy weight grey relational analysis (EW-GRA) is used as the DTR term to constrain the weight matrix and suppress irrelevant features. To reduce the accumulation of information loss during the reconstruction, an information enhancement layer is introduced, where the original inputs and the information of the hidden layers of previous DTR-EAE units are added to the follow-up DTR-EAE unit. Finally, in the regression process, the DTR term is used again to fully utilize depth features for quality prediction and prevent overfitting. Experimental verifications using the debutanizer column and thermal power plant are conducted to validate the effectiveness of the proposed modelling method.

关键词： autoencoder dynamic target regularization entropy weight grey relational analysis soft sensor

来源：评论

学校读者我要写书评

暂无评论

Baseline optimized autoencoder-based unsupervised anomaly detection in uncontrolled dynamic structural health monitoring

引用

STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL 2025年

作者： Yang, Kang Liu, Tianqi Yang, Zekun Zhou, Yang Tian, Zhihui Kim, Nam H. Harley, Joel B. Univ Florida Dept Elect & Comp Engn POB 116200216 Larsen Hall Gainesville FL 32608 USA Univ Florida Dept Comp & Informat Sci & Engn Gainesville FL USA Univ Florida Dept Mech & Aerosp Engn Gainesville FL USA

autoencoder reconstruction-based unsupervised anomaly detection is widely used in structural health monitoring. However, these methods typically require training on historical data from healthy structures, collected under environmental conditions similar to the test data. This limits their practical use, as it demands a comprehensive dataset of historical guided waves gathered across various environmental and operational conditions. Additionally, these methods fail when the training data contain a significant portion of damage-induced guided waves, as the autoencoder may reconstruct damaged waves just as effectively as normal ones. To overcome these challenges, our anomaly detection model is trained directly on current measurements, eliminating the risk of environmental discrepancies between training and test data. Furthermore, our baseline optimization strategy biases the autoencoder toward reconstructing normal guided waves, enabling reliable anomaly detection even when a large proportion of the training data are damage-induced waves. Additionally, we present a strategy to enhance the model's practical performance by optimizing the weight factor for baseline loss and the baseline set size, based on guided wave reconstruction performance, without relying on damage labels. The effectiveness of this baseline-optimized autoencoder model, even when the training data contain significant damage-induced guided waves, is validated through measurements from 10 regions, each spanning 80 days of guided wave data collected under uncontrolled and dynamic environmental conditions.

关键词： Structural health monitoring guided waves anomaly detection autoencoder baseline optimization dynamic environmental conditions

来源：评论

学校读者我要写书评

暂无评论

iBhb-Lys: Identify lysine (3-hydroxybutyrylation sites using autoencoder feature representation and fuzzy SVM algorithm

引用

ANALYTICAL BIOCHEMISTRY 2025年 697卷 115715页

作者： Ju, Zhe Zhang, Qing-Bao Shenyang Aerosp Univ Coll Sci Shenyang 110136 Peoples R China

Lysine (3-hydroxybutyrylation (Kbhb) is newly discovered (3-hydroxybutyrylate-derived histone modification which has been associated with the pathogenesis of many human diseases. To further elucidate the biological significance and molecular mechanism of Kbhb, it is necessary to accurately identify the Kbhb sites from protein sequences. In this study, a novel computational model named iBhb-Lys is developed for the identification of Kbhb sites. Four types of features are combined to encode each Kbhb site as a 3266-dimensional feature vector. And the autoencoder network is used to reduce the dimensionality of feature space, due to the high dimensionality of the combined features. In addition, to effectively reduce the influence of noise and outlier on classification, a new fuzzy support vector machine algorithm is proposed by incorporating the density around the sample into the fuzzy membership function. As illustrated by independent test, the AUC value of iBhb-Lys has increased by 2.22 % compared to the existing predictor KbhbXG. Feature analysis shows that some amino acid composition features, such as the occurrence frequency of leucine and histidine residues around Kbhb sites, contribute profoundly to the identification of Kbhb sites. The conclusions drawn in this study may provide useful reference for studying the molecular mechanism of Kbhb.

关键词： Post-translational modification Feature extraction autoencoder Fuzzy support vector machine (3-hydroxybutyrylation

来源：评论

学校读者我要写书评

暂无评论

Intelligent evaluation method for identifying favorable shale oil areas based on improved stacked sparse autoencoder

引用

OIL SHALE 2025年第1期42卷 79-114页

作者： Xu, Rui Yan, Tie Sun, Shihui Qu, Jingyu Hou, Zhaokai Northeast Petr Univ Sanya Offshore Oil & Gas Res Inst Sanya 572025 Peoples R China Northeast Petr Univ Sch Petr Engn Daqing 163318 Peoples R China

Identifying and evaluating favorable areas is crucial for shale oil exploration and development, well-location deployment, and fracturing design. Traditional machine learning methods struggle to accurately extract the characteristics of favorable shale oil areas with limited labeled data, affecting accuracy and generalization. This study proposes an intelligent method for identifying favorable shale oil areas under semi-supervised learning (SSAEplus) to identify and evaluate favorable shale oil areas of the Qingshankou Formation in the Songliao Basin. The experimental results show that this method can effectively overcome the favorable area identification models' reliance on labeled data and can adaptively extract the characteristics of favorable shale oil areas without supervision. The accuracy of model identification is as high as 98.82%. Compared with other methods, the SSAE-plus yields higher accuracy and efficiency, while being more stable and generalizable. The SSAE-plus achieved over 95% accuracy in identifying favorable shale oil areas across six datasets. It has broad application prospects in identifying and evaluating favorable areas, and provides valuable theoretical insights for shale oil development and exploration well layout.

关键词： shale oil favorable area autoencoder batch normalization semi- supervised learning

来源：评论

学校读者我要写书评

暂无评论

Fault detection for high-speed train traction system using autoencoder-Fréchet inception distance

引用

MEASUREMENT SCIENCE AND TECHNOLOGY 2025年第4期36卷 046205-046205页

作者： Zhang, Cheng Lao, Yu-Yu Deng, Cheng-Long Li, Yuan Shenyang Univ Chem Technol Coll Sci Shenyang Liaoning Peoples R China Shenyang Univ Chem Technol Coll Comp Sci & Technol Shenyang Liaoning Peoples R China Shenyang Univ Chem Technol Coll Informat Engn Shenyang Liaoning Peoples R China

The detection of the process state in the traction system plays a crucial role in ensuring the safe operation of high-speed trains. However, the running data in the traction system exhibits nonlinear and dynamic nature, which leads to the failure of traditional detection methods. To address the above problems, a novel fault detection method based on autoencoder-Fr & eacute;chet inception distance (AE-FID) is proposed in this paper. Firstly, autoencoder (AE) is used for feature extraction to capture the inherent nonlinear structure in the data. Secondly, a novel fault detection statistic is developed by combining sliding window technology and Fr & eacute;chet inception distance (FID) to eliminate the influence of data dynamics. Finally, the proposed statistic is utilized to detect faults in the traction system, and the occurrence of faults is measured in terms of structural differences to mitigate the influence of the masking effects. The proposed method is evaluated through a numerical example that simulates dynamic nonlinear systems and the Traction Drive Control System-Fault Injection Benchmark simulation platform of a high-speed train. The experimental results demonstrate the effectiveness of the proposed method in fault detection for traction systems.

关键词： traction system fault detection autoencoder Fr & eacute chet inception distance sliding window

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：