检索结果-内蒙古大学图书馆

auto-encoder-based inverse characterization of Transport properties of acoustic foams

APPLIED ACOUSTICS 2025年 228卷

作者： Chen, Jianglong Xu, Yiqin Zhao, Xiaoliang Miao, Menghe Meng, Jiaguang Yu, Lingjie Zhi, Chao Xian Polytech Univ Sch Text Sci & Engn Xian 710048 Shaanxi Peoples R China Xian Polytech Univ State Key Lab Intelligent Text Mat & Prod Xian 710048 Shaanxi Peoples R China Univ Melbourne Dept Mech Engn Grattan St Parkville Vic 3010 Australia

The characterization of non-acoustic parameters is critically important for understanding the acoustic property and structural design of polyurethane (PU) foams. However, inverse characterization of acoustic PU foams through experiments and simulations often results in prolonged cycles and high resource wastage. To address the above issue, an innovative approach based on the auto-encoder (AE) was proposed in this paper. In the AE approach, the decoder module was utilized for the forward prediction part, while the encoder was used for the inverse characterization. A sample database of 96,730 data sets covering PU foams' sound absorption coefficients at 500-6000 Hz was established to train the AE model. To verify the effectiveness of the trained model, a comparative experiment with numerical simulations was firstly conducted. The results revealed that the coefficient of determination (R2) of forward prediction module surpasses 0.99, while the prediction time is significantly rapid, averaging 0.0005 s per sample, which is 1/22,000 of numerical simulation time. Another comparative experiment was conducted between the inverse characterization results of the machine learning model and the experimental data from real samples. The results showed that the average error of the characterization parameters (non-acoustic parameters and material thickness) is about 8.70 %. In summary, this study provides an intelligent inverse characterization method for targeted sound absorption of PU foams, with potential extensions to the inverse characterization of other acoustic porous materials.

关键词： Polyurethane foams Acoustic properties Non-acoustic parameters auto-encoder Inverse characterization

来源：评论

学校读者我要写书评

暂无评论

Efficient and robust de-speckling filter for ultrasound images based on an attentional auto-encoder

引用

SIGNAL IMAGE AND VIDEO PROCESSING 2025年第1期19卷 1-13页

作者： Ahmed, Anas Fouad Al Iraqia Univ Coll Engn Dept Elect Engn Baghdad 6029 Iraq

The existence of speckle noise in medical ultrasound (US) images greatly diminishes image quality and decreases diagnostic accuracy. The recent models for de-noising US images suffer from poor generalizability, a loss of texture information, and over-smoothing of the US images, especially at high noise levels. This study suggests an efficient and robust de-speckling filter for US images based on an attentional auto-encoder. This filter incorporates two distinct and efficient attention levels. The channel attention level (CAL) and large kernel attention level (KAL). These levels are integrated into the down-sampling and up-sampling steps, respectively. A skip link is added before the up-sampling stage to avoid the gradient vanishing during training. The performance of the introduced model was assessed using subjective visual evaluations and objective determination indices. The outcomes of the suggested approach on online databases achieved the minimum execution time (AET), the maximum values for structural similarity (SSIM), and peak signal-to-noise ratio (PSNR) indices over all levels of noise, exceeding other recent algorithms. For instance, at a high level of noise, the presented technique based on the breast database achieves the following mean values: 0.005, 0.81, and 23.71 dB for the AET, SSIM, and PSNR, respectively. Also, in clinical settings, the introduced model attained the highest values for the equivalent number of looks (ENL = 18.78), contrast-to-noise ratio (CNR = 5.12), and highest resolution (alpha similar to\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha<^>{\sim }$$\end{document} = 0.28). These findings prove that the presented architecture is more efficient and robust for rejecting noise while maintaining image details than other recent techniques.

关键词： Ultrasound images Speckle noise Filtering Attention strategy auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Gearbox fault identification using auto-encoder without training data from the damaged machine

引用

MEASUREMENT 2025年第PartA期242卷

作者： Pawlik, Pawel Kania, Konrad Przysucha, Bartosz AGH Univ Krakow Fac Mech Engn & Robot Dept Mech & Vibroacoust al A Mickiewicza 30 PL-30059 Krakow Poland Lublin Univ Technol Fac Management Dept Quantitat Methods Management ul Nadbystrzycka 38 PL-20618 Lublin Poland

Deep learning methods work well in machine diagnostics where operating conditions affect diagnostic signals. Classifiers are often used for fault identification, but these methods require training data sets measured for each fault. The solution to the lack of data is autoencoder-based network models, but these models can only detect, not identify faults. This article presents a new fault identification method based on auto-encoders (AE-based Fault Identification Technique AE-FIT) that does not require training data from the damaged machine. This method diagnoses pinion gearboxes operating under variable conditions (variable load, load-induced rotational speed, and temperature). The result of the technique is an interpretable diagnostic spectrum (AE-based Interpretable Order Spectrum AE- IOS). The method has been tested on two laboratory benches to detect misalignment, unbalance, and gearbox degradation. The damages introduced were used to validate a technique based on an auto-encoder trained only with data from undamaged machines.

关键词： Fault identification Degradation process Variable working conditions Neural networks Deep learning auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Hybrid model of convolutional auto-encoder and ellipse characteristic for unsupervised high impedance fault detection

引用

ELECTRIC POWER SYSTEMS RESEARCH 2025年 238卷

作者： Yang, Junjie Delinchant, Benoit Niyato, Dusit Hadjsaid, Nouredine CNRSCREATE LTD Singapore 138602 Singapore Univ Grenoble Alpes CNRS Grenoble INP G2Elab F-38031 Grenoble France Nanyang Technol Univ Coll Comp & Data Sci CCDS Singapore 639798 Singapore

High impedance faults (HIFs) present significant safety hazards such as fires and electric shock accidents, urging the development of reliable HIF detection approaches. Unfortunately, conventional protection relays often fail to detect HIFs due to their low fault currents. The growing utilization of sensors in distribution grids enables the monitoring of node signals, offering a promising way for accurate HIF detection. Nevertheless, challenges persist due to the scarcity of faulty data and the high costs associated with data collection. To tackle this challenge, we propose a hybrid artificial intelligence (AI) approach that combines the unsupervised learning capabilities of auto-encoders with the reliable physical knowledge of voltage waveform. This approach involves the transformation of node voltage signals into phase diagram curves, leveraging the ellipse characteristics of the curve to normalize the model's learning process. Performance evaluation conducted using the IEEE-34 and IEEE 13 nodes test feeders demonstrates its effectiveness in achieving highly accurate HIF detection across diverse fault conditions and locations. Comparative analysis against other unsupervised HIF detection methods reveals its superiority in terms of high detection accuracy and strong robustness to switching disturbances. The simulation data and codes relevant to this study are openly accessible, serving as a benchmark for research in HIF detection.

关键词： High impedance fault Unsupervised learning Fault detection PINN Phase diagram curve auto-encoder

来源：评论

学校读者我要写书评

暂无评论

Unsupervised feature learning using locality-preserved auto-encoder with complexity-invariant distance for intelligent fault diagnosis of machinery

引用

APPLIED INTELLIGENCE 2025年第6期55卷 1-14页

作者： Lu, Zhenghua Chu, Zhaobi Zhu, Min Dong, Xueping Hefei Univ Technol Sch Elect Engn & Automat Hefei 230009 Peoples R China

Unsupervised feature learning (UFL) has been recognized as a promising feature extractor in machinery fault diagnosis, where the auto-encoder is a very popular UFL framework. For the auto-encoder methods, however, it is still a great challenge to learn discriminative features from complex signals in an unsupervised manner. In this paper, a new UFL method named locality-preserved auto-encoder (LPAE) is proposed by explicitly designing a locality-preserved penalty term. Concretely, the penalty term constrains local geometry of samples in the original space to be well preserved in the reconstruction space, enabling more discriminative features to be learned accordingly. To better formulate this term, the complexity-invariant distance (CID) is employed to measure similarity between two mechanical signals so as to construct a reliable neighbor graph. On a rolling bearing dataset, experimental results verify that the proposed LPAE can learn sufficiently discriminative features from complex vibration signals collected from varying operating conditions, and achieves a remarkable and superior diagnosis performance over the existing advanced UFL methods. Moreover, the effectiveness of CID has been adequately validated by comparing with several other distance measurement methods. The proposed LPAE can be applied to the feature extraction stage of machinery fault diagnosis, which provides a potential solution for engineers to realize unsupervised learning of discriminative features.

关键词： Unsupervised feature learning Machinery fault diagnosis auto-encoder Locality Complexity-invariant distance

来源：评论

学校读者我要写书评

暂无评论

Generalization error guaranteed auto-encoder-based nonlinear model reduction for operator learning

引用

APPLIED AND COMPUTATIONAL HARMONIC ANALYSIS 2025年 74卷

作者： Liu, Hao Dahal, Biraj Lai, Rongjie Liao, Wenjing Hong Kong Baptist Univ Dept Math Hong Kong Peoples R China Georgia Inst Technol Sch Math Atlanta GA 30332 USA Purdue Univ Dept Math W Lafayette IN USA

Many physical processes in science and engineering are naturally represented by operators between infinite-dimensional function spaces. The problem of operator learning, in this context, seeks to extract these physical processes from empirical data, which is challenging due to the infinite or high dimensionality of data. An integral component in addressing this challenge is model reduction, which reduces both the data dimensionality and problem size. In this paper, we utilize low-dimensional nonlinear structures in model reduction by investigating auto-encoder- based Neural Network (AENet). AENet first learns the latent variables of the input data and then learns the transformation from these latent variables to corresponding output data. Our numerical experiments validate the ability of AENet to accurately learn the solution operator of nonlinear partial differential equations. Furthermore, we establish a mathematical and statistical estimation theory that analyzes the generalization error of AENet. Our theoretical framework shows that the sample complexity of training AENet is intricately tied to the intrinsic dimension of the modeled process, while also demonstrating the robustness of AENet to noise.

关键词： Deep learning theory Generalization error Model reduction auto-encoder Operator learning

来源：评论

学校读者我要写书评

暂无评论

Unsupervised feature selection using sparse manifold learning: auto-encoder approach

引用

INFORMATION PROCESSING & MANAGEMENT 2025年第1期62卷

作者： Moslemi, Amir Jamshidi, Mina Sunnybrook Res Inst SRI Phys Sci Res Platform Toronto ON Canada Seneca Polytech Sch Software Design & Data Sci Toronto ON Canada Grad Univ Adv Technol Dept Appl Math Kerman Iran

Feature selection techniques are widely being used as a preprocessing step to train machine learning algorithms to circumvent the curse of dimensionality, overfitting, and computation time challenges. Projection-based methods are frequently employed in feature selection, leveraging the extraction of linear relationships among features. The absence of nonlinear information extraction among features is notable in this context. While auto-encoder based techniques have recently gained traction for feature selection, their focus remains primarily on the encoding phase, as it is through this phase that the selected features are derived. The subtle point is that the performance of auto-encoder to obtain the most discriminative features is significantly affected by decoding phase. To address these challenges, in this paper, we proposed a novel feature selection based on auto-encoder to not only extracting nonlinear information among features but also decoding phase is regularized as well to enhance the performance of algorithm. In this study, we defined a new model of auto-encoder to preserve the topological information of reconstructed close to input data. To geometric structure of input data is preserved in projected space using Laplacian graph, and geometrical projected space is preserved in reconstructed space using a suitable term (abstract Laplacian graph of reconstructed data) in optimization problem. Preserving abstract Laplacian graph of reconstructed data close to Laplacian graph of input data affects the performance of feature selection and we experimentally showed this. Therefore, we show an effective approach to solve the objective of the corresponding problem. Since this approach can be mainly used for clustering aims, we conducted experiments on ten benchmark datasets and assessed our propped method based on clustering accuracy and normalized mutual information (NMI) metric. Our method obtained considerable superiority over recent state-of-the-art techniques

关键词： Feature selection auto-encoder Manifold structure

来源：评论

学校读者我要写书评

暂无评论

Heterogeneous Graph Transformer auto-encoder for multivariate time series forecasting

引用

COMPUTERS & ELECTRICAL ENGINEERING 2025年 122卷

作者： Ye, Hongjiang Sun, Ying Gao, Yu Xu, Feiyi Qi, Jin Nanjing Univ Posts & Telecommun Sch Internet Things Nanjing 210003 Peoples R China Nanjing Univ Posts & Telecommun Sch Comp Sci Nanjing 210023 Peoples R China Nanjing Univ Posts & Telecommun Sch Automat & Artificial Intelligence Nanjing 210023 Jiangsu Peoples R China NanJing Pharmaceut Co Ltd Digital Innovat Dept Nanjing 210012 Peoples R China

Multivariate time series (MTS) forecasting is crucial for optimizing intelligent applications. However, it is challenging because it requires the consideration of variable dependencies and temporal patterns based on complex relationships. Existing methods usually assume a single relationship between different variates, which results in the model learning a prominent shared temporal pattern. To address this problem, we innovatively use the heterogeneous graph to model multivariate time series. Specifically, we design the Heterogeneous Graph Transformer auto-encoder (HGTAE) to address this issue. In the data preprocessing stage, we construct MTS into the heterogeneous graph based on a priori knowledge of multiple relationships among different variates. Then, we design an adaptive module in HGTAE to embed different types of variates into the same dimension, which can uniformly represent the node features in the heterogeneous graph for subsequent graph processing and analysis. To learn the relationships and temporal information in the heterogeneous graph, we design an auto-encoder module in HGTAE. It learns correlations between nodes and their neighbors based on meta-relations, capturing temporal dependencies to reconstruct node attributes. We conduct link prediction and node prediction experiments on real-world heterogeneous graph datasets, and the results demonstrate the effectiveness of HGTAE in MTS forecasting tasks.

关键词： Heterogeneous graph learning Heterogeneous graph neural network Transformer auto-encoder Multivariate time series forecasting

来源：评论

学校读者我要写书评

暂无评论

Development of deep learning auto-encoder algorithms for predicting alcohol use in Korean adolescents based on cross-sectional data

引用

SOCIAL SCIENCE & MEDICINE 2025年 367卷 117690页

作者： Lee, Serim Ewha Womans Univ Dept Social Welf Seoul South Korea SUNY Albany Sch Publ Hlth Albany NY USA

Alcohol is a highly addictive substance, presenting significant global public health concerns, particularly among adolescents. Previous studies have been limited by traditional research methods, making it challenging to encompass diverse risk factors and automate screening or prediction of adolescents' alcohol use. This study aimed to develop prediction algorithms for adolescent alcohol use in South Korea using machine learning (ML) and deep learning (DL) models, and to identify important features. The study utilized a combination of DL (i.e., auto-encoder) and ML (i.e., Logistic regression, Ridge, LASSO, Elasticnet, Decision tree, Random forest, AdaBoost, and XGBoost) algorithms to develop the prediction models. It involves 41,239 Korean adolescents and 46 socio-ecological input variables based on cross-sectional data. The analysis revealed that the prediction algorithms had AUC scores ranging from 0.6325 to 0.7214. The feature importance analysis indicates that variables within the domains of sociodemographic characteristics, physical and mental health, behavioral problems, family factors, school factors, and social factors all play significant roles. The developed algorithms enable automatic and early identification of adolescent alcohol use within public health practice settings. By leveraging a comprehensive array of input variables, these methods surpass the limitations of traditional regression approaches, offering novel insights into the critical risk factors associated with alcohol use among Korean adolescents, thereby facilitating early and targeted prevention efforts.

关键词： Alcohol use Deep learning Machine learning auto-encoder Korean adolescent

来源：评论

学校读者我要写书评

暂无评论

A variational network for biomedical images denoising using bayesian model and auto-encoder

引用

BIOMEDICAL PHYSICS & ENGINEERING EXPRESS 2025年第1期11卷 2024 Dec 20页

作者： Kouanou, Aurelle Tchagna Karambal, Issa Gaba, Yae Tchito, Christian Tchapga Simo, Alain Marcel Dikande Takam, Clemence Alla Tchiotsop, Daniel Univ Buea Coll Technol Dept Comp Engn POB 63 Buea Cameroon InchTechs Solut Dept Training Res Dev & Innovat Yaounde Cameroon African Inst Math Sci AIMS Res & Innovat Ctr POB 6428 Kigali Rwanda Univ Buea Fac Sci Dept Phys Buea Cameroon Univ Dschang Unite Rech Automat & Informat Appl UR AIA IUT FV Bandjoun Bandjoun Cameroon

Background and Objective. auto-encoders have demonstrated outstanding performance in computer vision tasks such as biomedical imaging, including classification, segmentation, and denoising. Many of the current techniques for image denoising in biomedical applications involve training an autoencoder or convolutional neural network (CNN) using pairs of clean and noisy images. However, these approaches are not realistic because the autoencoder or CNN is trained on known noise and does not generalize well to new noisy distributions. This paper proposes a novel approach for biomedical image denoising using a variational network based on a Bayesian model and deep learning. Method. In this study, we aim to denoise biomedical images using a Bayesian approach. In our dataset, each image exhibited a same noise distribution. To achieve this, we first estimate the noise distribution based on Bayesian probability by calculating the posterior distributions, and then proceed with denoising. A loss function that combines the Bayesian prior and autoencoder objectives is used to train the variational network. The proposed method was tested on CT-Scan biomedical image datasets and compared with state-of-the-art denoising techniques. Results. The experimental results demonstrate that our method outperforms the existing methods in terms of denoising accuracy, visual quality, and computational efficiency. For instance, we obtained a PSNR of 39.18 dB and an SSIM of 0.9941 with noise intensity alpha = 10. Our approach can potentially improve the accuracy and reliability of biomedical image analysis, which can have significant implications for clinical diagnosis and treatment planning. Conclusion. The proposed method combines the advantages of both Bayesian modeling and variational network to effectively denoise biomedical images.

关键词： auto-encoder variational network bayesian model PSNR SSIM

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：