Missing data presents a crucial challenge in data analysis,especially in high-dimensional datasets,where missing data often leads to biased conclusions and degraded model performance.In this study,we present a novel a...Missing data presents a crucial challenge in data analysis,especially in high-dimensional datasets,where missing data often leads to biased conclusions and degraded model performance.In this study,we present a novel autoencoder-based imputation framework that integrates a composite loss function to enhance robustness and precision.The proposed loss combines(i)a guided,masked mean squared error focusing on missing entries;(ii)a noise-aware regularization term to improve resilience against data corruption;and(iii)a variance penalty to encourage expressive yet stable reconstructions.We evaluate the proposed model across four missingness mechanisms,such as Missing Completely at Random,Missing at Random,Missing Not at Random,and Missing Not at Random with quantile censorship,under systematically varied feature counts,sample sizes,and missingness ratios ranging from 5%to 60%.Four publicly available real-world datasets(Stroke Prediction,Pima Indians Diabetes,Cardiovascular Disease,and Framingham Heart Study)were used,and the obtained results show that our proposed model consistently outperforms baseline methods,including traditional and deep learning-based techniques.An ablation study reveals the additive value of each component in the loss function.Additionally,we assessed the downstream utility of imputed data through classification tasks,where datasets imputed by the proposed method yielded the highest receiver operating characteristic area under the curve scores across all scenarios.The model demonstrates strong scalability and robustness,improving performance with larger datasets and higher feature counts.These results underscore the capacity of the proposed method to produce not only numerically accurate but also semantically useful imputations,making it a promising solution for robust data recovery in clinical applications.展开更多
In the field of intelligent air combat,real-time and accurate recognition of within-visual-range(WVR)maneuver actions serves as the foundational cornerstone for constructing autonomous decision-making systems.However,...In the field of intelligent air combat,real-time and accurate recognition of within-visual-range(WVR)maneuver actions serves as the foundational cornerstone for constructing autonomous decision-making systems.However,existing methods face two major challenges:traditional feature engineering suffers from insufficient effective dimensionality in the feature space due to kinematic coupling,making it difficult to distinguish essential differences between maneuvers,while end-to-end deep learning models lack controllability in implicit feature learning and fail to model high-order long-range temporal dependencies.This paper proposes a trajectory feature pre-extraction method based on a Long-range Masked Autoencoder(LMAE),incorporating three key innovations:(1)Random Fragment High-ratio Masking(RFH-Mask),which enforces the model to learn long-range temporal correlations by masking 80%of trajectory data while retaining continuous fragments;(2)Kalman Filter-Guided Objective Function(KFG-OF),integrating trajectory continuity constraints to align the feature space with kinematic principles;and(3)Two-stage Decoupled Architecture,enabling efficient and controllable feature learning through unsupervised pre-training and frozen-feature transfer.Experimental results demonstrate that LMAE significantly improves the average recognition accuracy for 20-class maneuvers compared to traditional end-to-end models,while significantly accelerating convergence speed.The contributions of this work lie in:introducing high-masking-rate autoencoders into low-informationdensity trajectory analysis,proposing a feature engineering framework with enhanced controllability and efficiency,and providing a novel technical pathway for intelligent air combat decision-making systems.展开更多
Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redund...Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redundancy caused by local structural similarities in 3D objects,which limits the performance of sampling.To address this issue,this paper introduces a novel task-oriented point cloud masked autoencoder-based sampling network(Point-MASNet),inspired by the masked autoencoder mechanism.Point-MASNet employs a voxel-based random non-overlapping masking strategy,which allows the model to selectively learn and capture distinctive local structural features from the input data.This approach effectively mitigates redundancy and enhances the representativeness of the sampled subset.In addition,we propose a lightweight,symmetrically structured keypoint reconstruction network,designed as an autoencoder.This network is optimized to efficiently extract latent features while enabling refined reconstructions.Extensive experiments demonstrate that Point-MASNet achieves competitive sampling performance across classification,registration,and reconstruction tasks.展开更多
Osteoporosis is a major cause of bone fracture and can be characterised by both mass loss and microstructure deterioration of the bone.The modern way of osteoporosis assessment is through the measurement of bone miner...Osteoporosis is a major cause of bone fracture and can be characterised by both mass loss and microstructure deterioration of the bone.The modern way of osteoporosis assessment is through the measurement of bone mineral density,which is not able to unveil the pathological condition from the mesoscale aspect.To obtain mesoscale information from computed tomography(CT),the super-resolution(SR)approach for volumetric imaging data is required.A deep learning model AESR3D is proposed to recover high-resolution(HR)Micro-CT from low-resolution Micro-CT and implement an unsupervised segmentation for better trabecular observation and measurement.A new regularisation overcomplete autoencoder framework for the SR task is proposed and theoretically analysed.The best performance is achieved on structural similarity measure of trabecular CT SR task compared with the state-of-the-art models in both natural and medical image SR tasks.The HR and SR images show a high correlation(r=0.996,intraclass correlation coefficients=0.917)on trabecular bone morphological indicators.The results also prove the effectiveness of our regularisation framework when training a large capacity model.展开更多
Electronic nose and thermal images are effective ways to diagnose the presence of gases in real-time realtime.Multimodal fusion of these modalities can result in the development of highly accurate diagnostic systems.T...Electronic nose and thermal images are effective ways to diagnose the presence of gases in real-time realtime.Multimodal fusion of these modalities can result in the development of highly accurate diagnostic systems.The low-cost thermal imaging software produces low-resolution thermal images in grayscale format,hence necessitating methods for improving the resolution and colorizing the images.The objective of this paper is to develop and train a super-resolution generative adversarial network for improving the resolution of the thermal images,followed by a sparse autoencoder for colorization of thermal images and amultimodal convolutional neural network for gas detection using electronic nose and thermal images.The dataset used comprises 6400 thermal images and electronic nose measurements for four classes.A multimodal Convolutional Neural Network(CNN)comprising an EfficientNetB2 pre-trainedmodel was developed using both early and late feature fusion.The Super Resolution Generative Adversarial Network(SRGAN)model was developed and trained on low and high-resolution thermal images.Asparse autoencoder was trained on the grayscale and colorized thermal images.The SRGAN was trained on lowand high-resolution thermal images,achieving a Structural Similarity Index(SSIM)of 90.28,a Peak Signal-to-Noise Ratio(PSNR)of 68.74,and a Mean Absolute Error(MAE)of 0.066.The autoencoder model produced an MAE of 0.035,a Mean Squared Error(MSE)of 0.006,and a Root Mean Squared Error(RMSE)of 0.0705.The multimodal CNN,trained on these images and electronic nose measurements using both early and late fusion techniques,achieved accuracies of 97.89% and 98.55%,respectively.Hence,the proposed framework can be of great aid for the integration with low-cost software to generate high quality thermal camera images and highly accurate detection of gases in real-time.展开更多
High-speed railway bridges are essential components of any railway transportation system that should keep adequate levels of serviceability and safety.In this context,drive-by methodologies have emerged as a feasible ...High-speed railway bridges are essential components of any railway transportation system that should keep adequate levels of serviceability and safety.In this context,drive-by methodologies have emerged as a feasible and cost-effective monitor-ing solution for detecting damage on railway bridges while minimizing train operation interruptions.Moreover,integrating advanced sensor technologies and machine learning algorithms has significantly enhanced structural health monitoring(SHM)for bridges.Despite being increasingly used in traditional SHM applications,studies using autoencoders within drive-by methodologies are rare,especially in the railway field.This study presents a novel approach for drive-by damage detection in HSR bridges.The methodology relies on acceleration records collected from multiple bridge crossings by an operational train equipped with onboard sensors.Log-Mel spectrogram features derived from the acceleration records are used together with sparse autoencoders for computing statistical distribution-based damage indexes.Numerical simulations were performed on a 3D vehicle-track-bridge interaction system model implemented in Matlab to evaluate the robustness and effectiveness of the proposed approach,considering several damage scenarios,vehicle speeds,and environmental and operational variations,such as multiple track irregularities and varying measurement noise.The results show that the pro-posed approach can successfully detect damages,as well as characterize their severity,especially for very early-stage dam-ages.This demonstrates the high potential of applying Mel-frequency damage-sensitive features associated with machine learning algorithms in the drive-by condition assessment of high-speed railway bridges.展开更多
Intelligent sorting is an important prerequisite for the full quantitative consumption and harmless disposal of kitchen waste.The existing object detection method based on an ImageNet pre-trained model is an effective...Intelligent sorting is an important prerequisite for the full quantitative consumption and harmless disposal of kitchen waste.The existing object detection method based on an ImageNet pre-trained model is an effective way of sorting.Owing to significant domain gaps between natural images and kitchen waste images,it is difficult to reflect the characteristics of diverse scales and dense distribution in kitchen waste based on an ImageNet pre-trained model,leading to poor generalisation.In this article,the authors propose the first pre-trained model for kitchen waste sorting called KitWaSor,which combines both contrastive learning(CL)and masked image modelling(MIM)through self-supervised learning(SSL).First,to address the issue of diverse scales,the authors propose a mixed masking strategy by introducing an incomplete masking branch based on the original random masking branch.It prevents the complete loss of small-scale objects while avoiding excessive leakage of large-scale object pixels.Second,to address the issue of dense distribution,the authors introduce semantic consistency constraints on the basis of the mixed masking strategy.That is,object semantic reasoning is performed through semantic consistency constraints to compensate for the lack of contextual information.To train KitWaSor,the authors construct the first million-level kitchen waste dataset across seasonal and regional distributions,named KWD-Million.Extensive experiments show that KitWaSor achieves state-of-the-art(SOTA)performance on the two most relevant downstream tasks for kitchen waste sorting(i.e.image classification and object detection),demonstrating the effectiveness of the proposed KitWaSor.展开更多
It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and externa...It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and external factors due to the complex conditions.Given that the existing models fail to take into account all the factors and accurate prediction of the multiple time series simultaneously is difficult using these models,this study proposed an improved prediction model through the autoencoder fused long-and short-term time-series network driven by the mass number of monitoring data.Then,the proposed model was formalized on multiple time series of strain monitoring data.Also,the discussion analysis with a classical baseline and an ablation experiment was conducted to verify the effectiveness of the prediction model.As the results indicate,the proposed model shows obvious superiority in predicting the future mechanical behaviors of structures.As a case study,the presented model was applied to the Nanjing Dinghuaimen tunnel to predict the stain variation on a different time scale in the future.展开更多
Predicting cross-immunity between viral strains is vital for public health surveillance and vaccine development.Traditional neural network methods,such as BiLSTM,could be ineffective due to the lack of lab data for mo...Predicting cross-immunity between viral strains is vital for public health surveillance and vaccine development.Traditional neural network methods,such as BiLSTM,could be ineffective due to the lack of lab data for model training and the overshadowing of crucial features within sequence concatenation.The current work proposes a less data-consuming model incorporating a pre-trained gene sequence model and a mutual information inference operator.Our methodology utilizes gene alignment and deduplication algorithms to preprocess gene sequences,enhancing the model’s capacity to discern and focus on distinctions among input gene pairs.The model,i.e.,DNA Pretrained Cross-Immunity Protection Inference model(DPCIPI),outperforms state-of-theart(SOTA)models in predicting hemagglutination inhibition titer from influenza viral gene sequences only.Improvement in binary cross-immunity prediction is 1.58%in F1,2.34%in precision,1.57%in recall,and 1.57%in Accuracy.For multilevel cross-immunity improvements,the improvement is 2.12%in F1,3.50%in precision,2.19%in recall,and 2.19%in Accuracy.Our study showcases the potential of pre-trained gene models to improve predictions of antigenic variation and cross-immunity.With expanding gene data and advancements in pre-trained models,this approach promises significant impacts on vaccine development and public health.展开更多
As cyber threats become increasingly sophisticated,Distributed Denial-of-Service(DDoS)attacks continue to pose a serious threat to network infrastructure,often disrupting critical services through overwhelming traffic...As cyber threats become increasingly sophisticated,Distributed Denial-of-Service(DDoS)attacks continue to pose a serious threat to network infrastructure,often disrupting critical services through overwhelming traffic.Although unsupervised anomaly detection using convolutional autoencoders(CAEs)has gained attention for its ability to model normal network behavior without requiring labeled data,conventional CAEs struggle to effectively distinguish between normal and attack traffic due to over-generalized reconstructions and naive anomaly scoring.To address these limitations,we propose CA-CAE,a novel anomaly detection framework designed to improve DDoS detection through asymmetric joint reconstruction learning and refined anomaly scoring.Our architecture connects two CAEs sequentially with asymmetric filter allocation,which amplifies reconstruction errors for anomalous data while preserving low errors for normal traffic.Additionally,we introduce a scoring mechanism that incorporates exponential decay weighting to emphasize recent anomalies and relative traffic volume adjustment to highlight highrisk instances,enabling more accurate and timely detection.We evaluate CA-CAE on a real-world network traffic dataset collected using Cisco NetFlow,containing over 190,000 normal instances and only 78 anomalous instances—an extremely imbalanced scenario(0.0004% anomalies).We validate the proposed framework through extensive experiments,including statistical tests and comparisons with baseline models.Despite this challenge,our method achieves significant improvement,increasing the F1-score from 0.515 obtained by the baseline CAE to 0.934,and outperforming other models.These results demonstrate the effectiveness,scalability,and practicality of CA-CAE for unsupervised DDoS detection in realistic network environments.By combining lightweight model architecture with a domain-aware scoring strategy,our framework provides a robust solution for early detection of DDoS attacks without relying on labeled attack data.展开更多
Deep neural networks provide accurate results for most applications.However,they need a big dataset to train properly.Providing a big dataset is a significant challenge in most applications.Image augmentation refers t...Deep neural networks provide accurate results for most applications.However,they need a big dataset to train properly.Providing a big dataset is a significant challenge in most applications.Image augmentation refers to techniques that increase the amount of image data.Common operations for image augmentation include changes in illumination,rotation,contrast,size,viewing angle,and others.Recently,Generative Adversarial Networks(GANs)have been employed for image generation.However,like image augmentation methods,GAN approaches can only generate images that are similar to the original images.Therefore,they also cannot generate new classes of data.Texture images presentmore challenges than general images,and generating textures is more complex than creating other types of images.This study proposes a gradient-based deep neural network method that generates a new class of texture.It is possible to rapidly generate new classes of textures using different kernels from pre-trained deep networks.After generating new textures for each class,the number of textures increases through image augmentation.During this process,several techniques are proposed to automatically remove incomplete and similar textures that are created.The proposed method is faster than some well-known generative networks by around 4 to 10 times.In addition,the quality of the generated textures surpasses that of these networks.The proposed method can generate textures that surpass those of someGANs and parametric models in certain image qualitymetrics.It can provide a big texture dataset to train deep networks.A new big texture dataset is created artificially using the proposed method.This dataset is approximately 2 GB in size and comprises 30,000 textures,each 150×150 pixels in size,organized into 600 classes.It is uploaded to the Kaggle site and Google Drive.This dataset is called BigTex.Compared to other texture datasets,the proposed dataset is the largest and can serve as a comprehensive texture dataset for training more powerful deep neural networks and mitigating overfitting.展开更多
To predict the lithium-ion(Li-ion)battery degradation trajectory in the early phase,arranging the maintenance of battery energy storage systems is of great importance.However,under different operation conditions,Li-io...To predict the lithium-ion(Li-ion)battery degradation trajectory in the early phase,arranging the maintenance of battery energy storage systems is of great importance.However,under different operation conditions,Li-ion batteries present distinct degradation patterns,and it is challenging to capture negligible capacity fade in early cycles.Despite the data-driven method showing promising performance,insufficient data is still a big issue since the ageing experiments on the batteries are too slow and expensive.In this study,we proposed twin autoencoders integrated into a two-stage method to predict the early cycles'degradation trajectories.The two-stage method can properly predict the degradation from course to fine.The twin autoencoders serve as a feature extractor and a synthetic data generator,respectively.Ultimately,a learning procedure based on the long-short term memory(LSTM)network is designed to hybridize the learning process between the real and synthetic data.The performance of the proposed method is verified on three datasets,and the experimental results show that the proposed method can achieve accurate predictions compared to its competitors.展开更多
The emergence of Generative Adversarial Network(GAN)techniques has garnered significant attention from the research community for the development of Intrusion Detection Systems(IDS).However,conventional GAN-based IDS ...The emergence of Generative Adversarial Network(GAN)techniques has garnered significant attention from the research community for the development of Intrusion Detection Systems(IDS).However,conventional GAN-based IDS models face several challenges,including training instability,high computational costs,and system failures.To address these limitations,we propose a Hybrid Wasserstein GAN and Autoencoder Model(WGAN-AE)for intrusion detection.The proposed framework leverages the stability of WGAN and the feature extraction capabilities of the Autoencoder Model.The model was trained and evaluated using two recent benchmark datasets,5GNIDD and IDSIoT2024.When trained on the 5GNIDD dataset,the model achieved an average area under the precisionrecall curve is 99.8%using five-fold cross-validation and demonstrated a high detection accuracy of 97.35%when tested on independent test data.Additionally,the model is well-suited for deployment on resource-limited Internetof-Things(IoT)devices due to its ability to detect attacks within microseconds and its small memory footprint of 60.24 kB.Similarly,when trained on the IDSIoT2024 dataset,the model achieved an average PR-AUC of 94.09%and an attack detection accuracy of 97.35%on independent test data,with a memory requirement of 61.84 kB.Extensive simulation results demonstrate that the proposed hybrid model effectively addresses the shortcomings of traditional GAN-based IDS approaches in terms of detection accuracy,computational efficiency,and applicability to real-world IoT environments.展开更多
The aerial deployment method enables Unmanned Aerial Vehicles(UAVs)to be directly positioned at the required altitude for their mission.This method typically employs folding technology to improve loading efficiency,wi...The aerial deployment method enables Unmanned Aerial Vehicles(UAVs)to be directly positioned at the required altitude for their mission.This method typically employs folding technology to improve loading efficiency,with applications such as the gravity-only aerial deployment of high-aspect-ratio solar-powered UAVs,and aerial takeoff of fixed-wing drones in Mars research.However,the significant morphological changes during deployment are accompanied by strong nonlinear dynamic aerodynamic forces,which result in multiple degrees of freedom and an unstable character.This hinders the description and analysis of unknown dynamic behaviors,further leading to difficulties in the design of deployment strategies and flight control.To address this issue,this paper proposes an analysis method for dynamic behaviors during aerial deployment based on the Variational Autoencoder(VAE).Focusing on the gravity-only deployment problem of highaspect-ratio foldable-wing UAVs,the method encodes the multi-degree-of-freedom unstable motion signals into a low-dimensional feature space through a data-driven approach.By clustering in the feature space,this paper identifies and studies several dynamic behaviors during aerial deployment.The research presented in this paper offers a new method and perspective for feature extraction and analysis of complex and difficult-to-describe extreme flight dynamics,guiding the research on aerial deployment drones design and control strategies.展开更多
In the context of intelligent manufacturing,the modern hot strip mill process(HSMP)shows characteristics such as diversification of products,multi-specification batch production,and demand-oriented customization.These...In the context of intelligent manufacturing,the modern hot strip mill process(HSMP)shows characteristics such as diversification of products,multi-specification batch production,and demand-oriented customization.These characteristics pose significant challenges to ensuring process stability and consistency of product performance.Therefore,exploring the potential relationship between product performance and the production process,and developing a comprehensive performance evaluation method adapted to modern HSMP have become an urgent issue.A comprehensive performance evaluation method for HSMP by integrating multi-task learning and stacked performance-related autoencoder is proposed to solve the problems such as incomplete performance indicators(PIs)data,insufficient real-time acquisition requirements,and coupling of multiple PIs.First,according to the existing Chinese standards,a comprehensive performance evaluation grade strategy for strip steel is designed.The random forest model is established to predict and complete the parts of PIs data that could not be obtained in real-time.Second,a stacked performance-related autoencoder(SPAE)model is proposed to extract the deep features closely related to the product performance.Then,considering the correlation between PIs,the multi-task learning framework is introduced to output the subitem ratings and comprehensive product performance rating results of the strip steel online in real-time,where each task represents a subitem of comprehensive performance.Finally,the effectiveness of the method is verified on a real HSMP dataset,and the results show that the accuracy of the proposed method is as high as 94.8%,which is superior to the other comparative methods.展开更多
Deep Learning has been widely used to model soft sensors in modern industrial processes with nonlinear variables and uncertainty.Due to the outstanding ability for high-level feature extraction,stacked autoencoder(SAE...Deep Learning has been widely used to model soft sensors in modern industrial processes with nonlinear variables and uncertainty.Due to the outstanding ability for high-level feature extraction,stacked autoencoder(SAE)has been widely used to improve the model accuracy of soft sensors.However,with the increase of network layers,SAE may encounter serious information loss issues,which affect the modeling performance of soft sensors.Besides,there are typically very few labeled samples in the data set,which brings challenges to traditional neural networks to solve.In this paper,a multi-scale feature fused stacked autoencoder(MFF-SAE)is suggested for feature representation related to hierarchical output,where stacked autoencoder,mutual information(MI)and multi-scale feature fusion(MFF)strategies are integrated.Based on correlation analysis between output and input variables,critical hidden variables are extracted from the original variables in each autoencoder's input layer,which are correspondingly given varying weights.Besides,an integration strategy based on multi-scale feature fusion is adopted to mitigate the impact of information loss with the deepening of the network layers.Then,the MFF-SAE method is designed and stacked to form deep networks.Two practical industrial processes are utilized to evaluate the performance of MFF-SAE.Results from simulations indicate that in comparison to other cutting-edge techniques,the proposed method may considerably enhance the accuracy of soft sensor modeling,where the suggested method reduces the root mean square error(RMSE)by 71.8%,17.1%and 64.7%,15.1%,respectively.展开更多
Geochemical survey data are essential across Earth Science disciplines but are often affected by noise,which can obscure important geological signals and compromise subsequent prediction and interpretation.Quantifying...Geochemical survey data are essential across Earth Science disciplines but are often affected by noise,which can obscure important geological signals and compromise subsequent prediction and interpretation.Quantifying prediction uncertainty is hence crucial for robust geoscientific decision-making.This study proposes a novel deep learning framework,the Spatially Constrained Variational Autoencoder(SC-VAE),for denoising geochemical survey data with integrated uncertainty quantification.The SC-VAE incorporates spatial regularization,which enforces spatial coherence by modeling inter-sample relationships directly within the latent space.The performance of the SC-VAE was systematically evaluated against a standard Variational Autoencoder(VAE)using geochemical data from the gold polymetallic district in the northwestern part of Sichuan Province,China.Both models were optimized using Bayesian optimization,with objective functions specifically designed to maintain essential geostatistical characteristics.Evaluation metrics include variogram analysis,quantitative measures of spatial interpolation accuracy,visual assessment of denoised maps,and statistical analysis of data distributions,as well as decomposition of uncertainties.Results show that the SC-VAE achieves superior noise suppression and better preservation of spatial structure compared to the standard VAE,as demonstrated by a significant reduction in the variogram nugget effect and an increased partial sill.The SC-VAE produces denoised maps with clearer anomaly delineation and more regularized data distributions,effectively mitigating outliers and reducing kurtosis.Additionally,it delivers improved interpolation accuracy and spatially explicit uncertainty estimates,facilitating more reliable and interpretable assessments of prediction confidence.The SC-VAE framework thus provides a robust,geostatistically informed solution for enhancing the quality and interpretability of geochemical data,with broad applicability in mineral exploration,environmental geochemistry,and other Earth Science domains.展开更多
Virtual Power Plants(VPPs)are integral to modern energy systems,providing stability and reliability in the face of the inherent complexities and fluctuations of solar power data.Traditional anomaly detection methodolo...Virtual Power Plants(VPPs)are integral to modern energy systems,providing stability and reliability in the face of the inherent complexities and fluctuations of solar power data.Traditional anomaly detection methodologies often need to adequately handle these fluctuations from solar radiation and ambient temperature variations.We introduce the Memory-Enhanced Autoencoder with Adversarial Training(MemAAE)model to overcome these limitations,designed explicitly for robust anomaly detection in VPP environments.The MemAAE model integrates three principal components:an LSTM-based autoencoder that effectively captures temporal dynamics to distinguish between normal and anomalous behaviors,an adversarial training module that enhances system resilience across diverse operational scenarios,and a prediction module that aids the autoencoder during the reconstruction process,thereby facilitating precise anomaly identification.Furthermore,MemAAE features a memory mechanism that stores critical pattern information,mitigating overfitting,alongside a dynamic threshold adjustment mechanism that adapts detection thresholds in response to evolving operational conditions.Our empirical evaluation of the MemAAE model using real-world solar power data shows that the model outperforms other comparative models on both datasets.On the Sopan-Finder dataset,MemAAE has an accuracy of 99.17%and an F1-score of 95.79%,while on the Sunalab Faro PV 2017 dataset,it has an accuracy of 97.67%and an F1-score of 93.27%.Significant performance advantages have been achieved on both datasets.These results show that MemAAE model is an effective method for real-time anomaly detection in virtual power plants(VPPs),which can enhance robustness and adaptability to inherent variables in solar power generation.展开更多
We analyze the suitability of existing pre-trained transformer-based language models(PLMs)for abstractive text summarization on German technical healthcare texts.The study focuses on the multilingual capabilities of t...We analyze the suitability of existing pre-trained transformer-based language models(PLMs)for abstractive text summarization on German technical healthcare texts.The study focuses on the multilingual capabilities of these models and their ability to perform the task of abstractive text summarization in the healthcare field.The research hypothesis was that large language models could perform high-quality abstractive text summarization on German technical healthcare texts,even if the model is not specifically trained in that language.Through experiments,the research questions explore the performance of transformer language models in dealing with complex syntax constructs,the difference in performance between models trained in English and German,and the impact of translating the source text to English before conducting the summarization.We conducted an evaluation of four PLMs(GPT-3,a translation-based approach also utilizing GPT-3,a German language Model,and a domain-specific bio-medical model approach).The evaluation considered the informativeness using 3 types of metrics based on Recall-Oriented Understudy for Gisting Evaluation(ROUGE)and the quality of results which is manually evaluated considering 5 aspects.The results show that text summarization models could be used in the German healthcare domain and that domain-independent language models achieved the best results.The study proves that text summarization models can simplify the search for pre-existing German knowledge in various domains.展开更多
Drive-by techniques for bridge health monitoring have drawn increasing attention from researchers and practitioners,in the attempt to make bridge condition-based monitoring more cost-efficient.In this work,the authors...Drive-by techniques for bridge health monitoring have drawn increasing attention from researchers and practitioners,in the attempt to make bridge condition-based monitoring more cost-efficient.In this work,the authors propose a drive-by approach that takes advantage from bogie vertical accelerations to assess bridge health status.To do so,continuous wavelet transform is combined with multiple sparse autoencoders that allow for damage detection and localization across bridge span.According to authors’best knowledge,this is the first case in which an unsupervised technique,which relies on the use of sparse autoencoders,is used to localize damages.The bridge considered in this work is a Warren steel truss bridge,whose finite element model is referred to an actual structure,belonging to the Italian railway line.To investigate damage detection and localization performances,different operational variables are accounted for:train weight,forward speed and track irregularity evolution in time.Two configurations for the virtual measuring channels were investigated:as a result,better performances were obtained by exploiting the vertical accelerations of both the bogies of the leading coach instead of using only one single acceleration signal.展开更多
文摘Missing data presents a crucial challenge in data analysis,especially in high-dimensional datasets,where missing data often leads to biased conclusions and degraded model performance.In this study,we present a novel autoencoder-based imputation framework that integrates a composite loss function to enhance robustness and precision.The proposed loss combines(i)a guided,masked mean squared error focusing on missing entries;(ii)a noise-aware regularization term to improve resilience against data corruption;and(iii)a variance penalty to encourage expressive yet stable reconstructions.We evaluate the proposed model across four missingness mechanisms,such as Missing Completely at Random,Missing at Random,Missing Not at Random,and Missing Not at Random with quantile censorship,under systematically varied feature counts,sample sizes,and missingness ratios ranging from 5%to 60%.Four publicly available real-world datasets(Stroke Prediction,Pima Indians Diabetes,Cardiovascular Disease,and Framingham Heart Study)were used,and the obtained results show that our proposed model consistently outperforms baseline methods,including traditional and deep learning-based techniques.An ablation study reveals the additive value of each component in the loss function.Additionally,we assessed the downstream utility of imputed data through classification tasks,where datasets imputed by the proposed method yielded the highest receiver operating characteristic area under the curve scores across all scenarios.The model demonstrates strong scalability and robustness,improving performance with larger datasets and higher feature counts.These results underscore the capacity of the proposed method to produce not only numerically accurate but also semantically useful imputations,making it a promising solution for robust data recovery in clinical applications.
文摘In the field of intelligent air combat,real-time and accurate recognition of within-visual-range(WVR)maneuver actions serves as the foundational cornerstone for constructing autonomous decision-making systems.However,existing methods face two major challenges:traditional feature engineering suffers from insufficient effective dimensionality in the feature space due to kinematic coupling,making it difficult to distinguish essential differences between maneuvers,while end-to-end deep learning models lack controllability in implicit feature learning and fail to model high-order long-range temporal dependencies.This paper proposes a trajectory feature pre-extraction method based on a Long-range Masked Autoencoder(LMAE),incorporating three key innovations:(1)Random Fragment High-ratio Masking(RFH-Mask),which enforces the model to learn long-range temporal correlations by masking 80%of trajectory data while retaining continuous fragments;(2)Kalman Filter-Guided Objective Function(KFG-OF),integrating trajectory continuity constraints to align the feature space with kinematic principles;and(3)Two-stage Decoupled Architecture,enabling efficient and controllable feature learning through unsupervised pre-training and frozen-feature transfer.Experimental results demonstrate that LMAE significantly improves the average recognition accuracy for 20-class maneuvers compared to traditional end-to-end models,while significantly accelerating convergence speed.The contributions of this work lie in:introducing high-masking-rate autoencoders into low-informationdensity trajectory analysis,proposing a feature engineering framework with enhanced controllability and efficiency,and providing a novel technical pathway for intelligent air combat decision-making systems.
基金supported by the National Key Research and Development Program of China(2022YFB3103500)the National Natural Science Foundation of China(62473033,62571027)+1 种基金in part by the Beijing Natural Science Foundation(L231012)the State Scholarship Fund from the China Scholarship Council.
文摘Task-oriented point cloud sampling aims to select a representative subset from the input,tailored to specific application scenarios and task requirements.However,existing approaches rarely tackle the problem of redundancy caused by local structural similarities in 3D objects,which limits the performance of sampling.To address this issue,this paper introduces a novel task-oriented point cloud masked autoencoder-based sampling network(Point-MASNet),inspired by the masked autoencoder mechanism.Point-MASNet employs a voxel-based random non-overlapping masking strategy,which allows the model to selectively learn and capture distinctive local structural features from the input data.This approach effectively mitigates redundancy and enhances the representativeness of the sampled subset.In addition,we propose a lightweight,symmetrically structured keypoint reconstruction network,designed as an autoencoder.This network is optimized to efficiently extract latent features while enabling refined reconstructions.Extensive experiments demonstrate that Point-MASNet achieves competitive sampling performance across classification,registration,and reconstruction tasks.
基金Beijing Natural Science Foundation-Haidian original Innovation Joint Foundation,Grant/Award Number:L192016Joint Funds of the National Natural Science Foundation of China,Grant/Award Number:U21A20489+3 种基金National Natural Science Foundation of China,Grant/Award Number:62003330Shenzhen Fundamental Research Funds,Grant/Award Numbers:JCYJ20220818101608019,JCYJ20190807170407391,JCYJ20180507182415428Natural Science Foundation of Guangdong Province,Grant/Award Number:2019A1515011699Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems,Shenzhen Institute of Advanced Technology。
文摘Osteoporosis is a major cause of bone fracture and can be characterised by both mass loss and microstructure deterioration of the bone.The modern way of osteoporosis assessment is through the measurement of bone mineral density,which is not able to unveil the pathological condition from the mesoscale aspect.To obtain mesoscale information from computed tomography(CT),the super-resolution(SR)approach for volumetric imaging data is required.A deep learning model AESR3D is proposed to recover high-resolution(HR)Micro-CT from low-resolution Micro-CT and implement an unsupervised segmentation for better trabecular observation and measurement.A new regularisation overcomplete autoencoder framework for the SR task is proposed and theoretically analysed.The best performance is achieved on structural similarity measure of trabecular CT SR task compared with the state-of-the-art models in both natural and medical image SR tasks.The HR and SR images show a high correlation(r=0.996,intraclass correlation coefficients=0.917)on trabecular bone morphological indicators.The results also prove the effectiveness of our regularisation framework when training a large capacity model.
基金funded by the Centre for Advanced Modelling and Geospatial Information Systems(CAMGIS),Faculty of Engineering and IT,University of Technology Sydneysupported by the Researchers Supporting Project,King Saud University,Riyadh,Saudi Arabia,under Project RSP2025 R14.
文摘Electronic nose and thermal images are effective ways to diagnose the presence of gases in real-time realtime.Multimodal fusion of these modalities can result in the development of highly accurate diagnostic systems.The low-cost thermal imaging software produces low-resolution thermal images in grayscale format,hence necessitating methods for improving the resolution and colorizing the images.The objective of this paper is to develop and train a super-resolution generative adversarial network for improving the resolution of the thermal images,followed by a sparse autoencoder for colorization of thermal images and amultimodal convolutional neural network for gas detection using electronic nose and thermal images.The dataset used comprises 6400 thermal images and electronic nose measurements for four classes.A multimodal Convolutional Neural Network(CNN)comprising an EfficientNetB2 pre-trainedmodel was developed using both early and late feature fusion.The Super Resolution Generative Adversarial Network(SRGAN)model was developed and trained on low and high-resolution thermal images.Asparse autoencoder was trained on the grayscale and colorized thermal images.The SRGAN was trained on lowand high-resolution thermal images,achieving a Structural Similarity Index(SSIM)of 90.28,a Peak Signal-to-Noise Ratio(PSNR)of 68.74,and a Mean Absolute Error(MAE)of 0.066.The autoencoder model produced an MAE of 0.035,a Mean Squared Error(MSE)of 0.006,and a Root Mean Squared Error(RMSE)of 0.0705.The multimodal CNN,trained on these images and electronic nose measurements using both early and late fusion techniques,achieved accuracies of 97.89% and 98.55%,respectively.Hence,the proposed framework can be of great aid for the integration with low-cost software to generate high quality thermal camera images and highly accurate detection of gases in real-time.
基金support of CNPq(Brazilian Ministry of Science and Technology Agency),of CAPES(Higher Education Improvement Agency),of FAPESP(São Paulo Research Foundation)under grant#2022/13045-1,of VALE Catedra Under Rail and of Base Funding-UIDB/04708/2020Programmatic Funding-UIDP/04708/2020 of the CONSTRUCT-“Instituto de I&D em Estruturas e Construções”.
文摘High-speed railway bridges are essential components of any railway transportation system that should keep adequate levels of serviceability and safety.In this context,drive-by methodologies have emerged as a feasible and cost-effective monitor-ing solution for detecting damage on railway bridges while minimizing train operation interruptions.Moreover,integrating advanced sensor technologies and machine learning algorithms has significantly enhanced structural health monitoring(SHM)for bridges.Despite being increasingly used in traditional SHM applications,studies using autoencoders within drive-by methodologies are rare,especially in the railway field.This study presents a novel approach for drive-by damage detection in HSR bridges.The methodology relies on acceleration records collected from multiple bridge crossings by an operational train equipped with onboard sensors.Log-Mel spectrogram features derived from the acceleration records are used together with sparse autoencoders for computing statistical distribution-based damage indexes.Numerical simulations were performed on a 3D vehicle-track-bridge interaction system model implemented in Matlab to evaluate the robustness and effectiveness of the proposed approach,considering several damage scenarios,vehicle speeds,and environmental and operational variations,such as multiple track irregularities and varying measurement noise.The results show that the pro-posed approach can successfully detect damages,as well as characterize their severity,especially for very early-stage dam-ages.This demonstrates the high potential of applying Mel-frequency damage-sensitive features associated with machine learning algorithms in the drive-by condition assessment of high-speed railway bridges.
基金National Key Research and Development Program of China,Grant/Award Number:2021YFC1910402。
文摘Intelligent sorting is an important prerequisite for the full quantitative consumption and harmless disposal of kitchen waste.The existing object detection method based on an ImageNet pre-trained model is an effective way of sorting.Owing to significant domain gaps between natural images and kitchen waste images,it is difficult to reflect the characteristics of diverse scales and dense distribution in kitchen waste based on an ImageNet pre-trained model,leading to poor generalisation.In this article,the authors propose the first pre-trained model for kitchen waste sorting called KitWaSor,which combines both contrastive learning(CL)and masked image modelling(MIM)through self-supervised learning(SSL).First,to address the issue of diverse scales,the authors propose a mixed masking strategy by introducing an incomplete masking branch based on the original random masking branch.It prevents the complete loss of small-scale objects while avoiding excessive leakage of large-scale object pixels.Second,to address the issue of dense distribution,the authors introduce semantic consistency constraints on the basis of the mixed masking strategy.That is,object semantic reasoning is performed through semantic consistency constraints to compensate for the lack of contextual information.To train KitWaSor,the authors construct the first million-level kitchen waste dataset across seasonal and regional distributions,named KWD-Million.Extensive experiments show that KitWaSor achieves state-of-the-art(SOTA)performance on the two most relevant downstream tasks for kitchen waste sorting(i.e.image classification and object detection),demonstrating the effectiveness of the proposed KitWaSor.
基金National Key Research and Development Program of China,Grant/Award Number:2018YFB2101003National Natural Science Foundation of China,Grant/Award Numbers:51991395,U1806226,51778033,51822802,71901011,U1811463,51991391Science and Technology Major Project of Beijing,Grant/Award Number:Z191100002519012。
文摘It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and external factors due to the complex conditions.Given that the existing models fail to take into account all the factors and accurate prediction of the multiple time series simultaneously is difficult using these models,this study proposed an improved prediction model through the autoencoder fused long-and short-term time-series network driven by the mass number of monitoring data.Then,the proposed model was formalized on multiple time series of strain monitoring data.Also,the discussion analysis with a classical baseline and an ablation experiment was conducted to verify the effectiveness of the prediction model.As the results indicate,the proposed model shows obvious superiority in predicting the future mechanical behaviors of structures.As a case study,the presented model was applied to the Nanjing Dinghuaimen tunnel to predict the stain variation on a different time scale in the future.
基金supported by the Bill & Melinda Gates Foundation and the Minderoo Foundation
文摘Predicting cross-immunity between viral strains is vital for public health surveillance and vaccine development.Traditional neural network methods,such as BiLSTM,could be ineffective due to the lack of lab data for model training and the overshadowing of crucial features within sequence concatenation.The current work proposes a less data-consuming model incorporating a pre-trained gene sequence model and a mutual information inference operator.Our methodology utilizes gene alignment and deduplication algorithms to preprocess gene sequences,enhancing the model’s capacity to discern and focus on distinctions among input gene pairs.The model,i.e.,DNA Pretrained Cross-Immunity Protection Inference model(DPCIPI),outperforms state-of-theart(SOTA)models in predicting hemagglutination inhibition titer from influenza viral gene sequences only.Improvement in binary cross-immunity prediction is 1.58%in F1,2.34%in precision,1.57%in recall,and 1.57%in Accuracy.For multilevel cross-immunity improvements,the improvement is 2.12%in F1,3.50%in precision,2.19%in recall,and 2.19%in Accuracy.Our study showcases the potential of pre-trained gene models to improve predictions of antigenic variation and cross-immunity.With expanding gene data and advancements in pre-trained models,this approach promises significant impacts on vaccine development and public health.
基金supported by Korea National University of Transportation Industry-Academy Cooperation Foundation in 2024.
文摘As cyber threats become increasingly sophisticated,Distributed Denial-of-Service(DDoS)attacks continue to pose a serious threat to network infrastructure,often disrupting critical services through overwhelming traffic.Although unsupervised anomaly detection using convolutional autoencoders(CAEs)has gained attention for its ability to model normal network behavior without requiring labeled data,conventional CAEs struggle to effectively distinguish between normal and attack traffic due to over-generalized reconstructions and naive anomaly scoring.To address these limitations,we propose CA-CAE,a novel anomaly detection framework designed to improve DDoS detection through asymmetric joint reconstruction learning and refined anomaly scoring.Our architecture connects two CAEs sequentially with asymmetric filter allocation,which amplifies reconstruction errors for anomalous data while preserving low errors for normal traffic.Additionally,we introduce a scoring mechanism that incorporates exponential decay weighting to emphasize recent anomalies and relative traffic volume adjustment to highlight highrisk instances,enabling more accurate and timely detection.We evaluate CA-CAE on a real-world network traffic dataset collected using Cisco NetFlow,containing over 190,000 normal instances and only 78 anomalous instances—an extremely imbalanced scenario(0.0004% anomalies).We validate the proposed framework through extensive experiments,including statistical tests and comparisons with baseline models.Despite this challenge,our method achieves significant improvement,increasing the F1-score from 0.515 obtained by the baseline CAE to 0.934,and outperforming other models.These results demonstrate the effectiveness,scalability,and practicality of CA-CAE for unsupervised DDoS detection in realistic network environments.By combining lightweight model architecture with a domain-aware scoring strategy,our framework provides a robust solution for early detection of DDoS attacks without relying on labeled attack data.
基金supported via funding from Prince Sattam bin Abdulaziz University(PSAU/2025/R/1446)Princess Nourah bint Abdulrahman University(PNURSP2025R300)Prince Sultan University.
文摘Deep neural networks provide accurate results for most applications.However,they need a big dataset to train properly.Providing a big dataset is a significant challenge in most applications.Image augmentation refers to techniques that increase the amount of image data.Common operations for image augmentation include changes in illumination,rotation,contrast,size,viewing angle,and others.Recently,Generative Adversarial Networks(GANs)have been employed for image generation.However,like image augmentation methods,GAN approaches can only generate images that are similar to the original images.Therefore,they also cannot generate new classes of data.Texture images presentmore challenges than general images,and generating textures is more complex than creating other types of images.This study proposes a gradient-based deep neural network method that generates a new class of texture.It is possible to rapidly generate new classes of textures using different kernels from pre-trained deep networks.After generating new textures for each class,the number of textures increases through image augmentation.During this process,several techniques are proposed to automatically remove incomplete and similar textures that are created.The proposed method is faster than some well-known generative networks by around 4 to 10 times.In addition,the quality of the generated textures surpasses that of these networks.The proposed method can generate textures that surpass those of someGANs and parametric models in certain image qualitymetrics.It can provide a big texture dataset to train deep networks.A new big texture dataset is created artificially using the proposed method.This dataset is approximately 2 GB in size and comprises 30,000 textures,each 150×150 pixels in size,organized into 600 classes.It is uploaded to the Kaggle site and Google Drive.This dataset is called BigTex.Compared to other texture datasets,the proposed dataset is the largest and can serve as a comprehensive texture dataset for training more powerful deep neural networks and mitigating overfitting.
基金financially supported by the National Natural Science Foundation of China under Grant 62372369,52107229,62272383the Key Research and Development Program of Shaanxi Province(2024GX-YBXM-442)Natural Science Basic Research Program of Shaanxi Province(2024JC-YBMS-477)。
文摘To predict the lithium-ion(Li-ion)battery degradation trajectory in the early phase,arranging the maintenance of battery energy storage systems is of great importance.However,under different operation conditions,Li-ion batteries present distinct degradation patterns,and it is challenging to capture negligible capacity fade in early cycles.Despite the data-driven method showing promising performance,insufficient data is still a big issue since the ageing experiments on the batteries are too slow and expensive.In this study,we proposed twin autoencoders integrated into a two-stage method to predict the early cycles'degradation trajectories.The two-stage method can properly predict the degradation from course to fine.The twin autoencoders serve as a feature extractor and a synthetic data generator,respectively.Ultimately,a learning procedure based on the long-short term memory(LSTM)network is designed to hybridize the learning process between the real and synthetic data.The performance of the proposed method is verified on three datasets,and the experimental results show that the proposed method can achieve accurate predictions compared to its competitors.
基金the Deanship of Research and Graduate Studies at King Khalid University for funding this work through Large Group Project under grant number(RGP.2/245/46)funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R760)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.The research team thanks the Deanship of Graduate Studies and Scientific Research at Najran University for supporting the research project through the Nama’a program,with the project code NU/GP/SERC/13/352-1。
文摘The emergence of Generative Adversarial Network(GAN)techniques has garnered significant attention from the research community for the development of Intrusion Detection Systems(IDS).However,conventional GAN-based IDS models face several challenges,including training instability,high computational costs,and system failures.To address these limitations,we propose a Hybrid Wasserstein GAN and Autoencoder Model(WGAN-AE)for intrusion detection.The proposed framework leverages the stability of WGAN and the feature extraction capabilities of the Autoencoder Model.The model was trained and evaluated using two recent benchmark datasets,5GNIDD and IDSIoT2024.When trained on the 5GNIDD dataset,the model achieved an average area under the precisionrecall curve is 99.8%using five-fold cross-validation and demonstrated a high detection accuracy of 97.35%when tested on independent test data.Additionally,the model is well-suited for deployment on resource-limited Internetof-Things(IoT)devices due to its ability to detect attacks within microseconds and its small memory footprint of 60.24 kB.Similarly,when trained on the IDSIoT2024 dataset,the model achieved an average PR-AUC of 94.09%and an attack detection accuracy of 97.35%on independent test data,with a memory requirement of 61.84 kB.Extensive simulation results demonstrate that the proposed hybrid model effectively addresses the shortcomings of traditional GAN-based IDS approaches in terms of detection accuracy,computational efficiency,and applicability to real-world IoT environments.
基金co-supported by the Natural Science Basic Research Program of Shaanxi,China(No.2023-JC-QN-0043)the ND Basic Research Funds,China(No.G2022WD).
文摘The aerial deployment method enables Unmanned Aerial Vehicles(UAVs)to be directly positioned at the required altitude for their mission.This method typically employs folding technology to improve loading efficiency,with applications such as the gravity-only aerial deployment of high-aspect-ratio solar-powered UAVs,and aerial takeoff of fixed-wing drones in Mars research.However,the significant morphological changes during deployment are accompanied by strong nonlinear dynamic aerodynamic forces,which result in multiple degrees of freedom and an unstable character.This hinders the description and analysis of unknown dynamic behaviors,further leading to difficulties in the design of deployment strategies and flight control.To address this issue,this paper proposes an analysis method for dynamic behaviors during aerial deployment based on the Variational Autoencoder(VAE).Focusing on the gravity-only deployment problem of highaspect-ratio foldable-wing UAVs,the method encodes the multi-degree-of-freedom unstable motion signals into a low-dimensional feature space through a data-driven approach.By clustering in the feature space,this paper identifies and studies several dynamic behaviors during aerial deployment.The research presented in this paper offers a new method and perspective for feature extraction and analysis of complex and difficult-to-describe extreme flight dynamics,guiding the research on aerial deployment drones design and control strategies.
基金supported by the National Natural Science Foundation of China(NSFC)under Grants(Nos.U21A20483,62373040 and 62273031).
文摘In the context of intelligent manufacturing,the modern hot strip mill process(HSMP)shows characteristics such as diversification of products,multi-specification batch production,and demand-oriented customization.These characteristics pose significant challenges to ensuring process stability and consistency of product performance.Therefore,exploring the potential relationship between product performance and the production process,and developing a comprehensive performance evaluation method adapted to modern HSMP have become an urgent issue.A comprehensive performance evaluation method for HSMP by integrating multi-task learning and stacked performance-related autoencoder is proposed to solve the problems such as incomplete performance indicators(PIs)data,insufficient real-time acquisition requirements,and coupling of multiple PIs.First,according to the existing Chinese standards,a comprehensive performance evaluation grade strategy for strip steel is designed.The random forest model is established to predict and complete the parts of PIs data that could not be obtained in real-time.Second,a stacked performance-related autoencoder(SPAE)model is proposed to extract the deep features closely related to the product performance.Then,considering the correlation between PIs,the multi-task learning framework is introduced to output the subitem ratings and comprehensive product performance rating results of the strip steel online in real-time,where each task represents a subitem of comprehensive performance.Finally,the effectiveness of the method is verified on a real HSMP dataset,and the results show that the accuracy of the proposed method is as high as 94.8%,which is superior to the other comparative methods.
基金supported by the National Key Research and Development Program of China(2023YFB3307800)National Natural Science Foundation of China(62394343,62373155)+2 种基金Major Science and Technology Project of Xinjiang(No.2022A01006-4)State Key Laboratory of Industrial Control Technology,China(Grant No.ICT2024A26)Fundamental Research Funds for the Central Universities.
文摘Deep Learning has been widely used to model soft sensors in modern industrial processes with nonlinear variables and uncertainty.Due to the outstanding ability for high-level feature extraction,stacked autoencoder(SAE)has been widely used to improve the model accuracy of soft sensors.However,with the increase of network layers,SAE may encounter serious information loss issues,which affect the modeling performance of soft sensors.Besides,there are typically very few labeled samples in the data set,which brings challenges to traditional neural networks to solve.In this paper,a multi-scale feature fused stacked autoencoder(MFF-SAE)is suggested for feature representation related to hierarchical output,where stacked autoencoder,mutual information(MI)and multi-scale feature fusion(MFF)strategies are integrated.Based on correlation analysis between output and input variables,critical hidden variables are extracted from the original variables in each autoencoder's input layer,which are correspondingly given varying weights.Besides,an integration strategy based on multi-scale feature fusion is adopted to mitigate the impact of information loss with the deepening of the network layers.Then,the MFF-SAE method is designed and stacked to form deep networks.Two practical industrial processes are utilized to evaluate the performance of MFF-SAE.Results from simulations indicate that in comparison to other cutting-edge techniques,the proposed method may considerably enhance the accuracy of soft sensor modeling,where the suggested method reduces the root mean square error(RMSE)by 71.8%,17.1%and 64.7%,15.1%,respectively.
基金supported by the National Natural Science Foundation of China(Nos.42530801,42425208)the Natural Science Foundation of Hubei Province(China)(No.2023AFA001)+1 种基金the MOST Special Fund from State Key Laboratory of Geological Processes and Mineral Resources,China University of Geosciences(No.MSFGPMR2025-401)the China Scholarship Council(No.202306410181)。
文摘Geochemical survey data are essential across Earth Science disciplines but are often affected by noise,which can obscure important geological signals and compromise subsequent prediction and interpretation.Quantifying prediction uncertainty is hence crucial for robust geoscientific decision-making.This study proposes a novel deep learning framework,the Spatially Constrained Variational Autoencoder(SC-VAE),for denoising geochemical survey data with integrated uncertainty quantification.The SC-VAE incorporates spatial regularization,which enforces spatial coherence by modeling inter-sample relationships directly within the latent space.The performance of the SC-VAE was systematically evaluated against a standard Variational Autoencoder(VAE)using geochemical data from the gold polymetallic district in the northwestern part of Sichuan Province,China.Both models were optimized using Bayesian optimization,with objective functions specifically designed to maintain essential geostatistical characteristics.Evaluation metrics include variogram analysis,quantitative measures of spatial interpolation accuracy,visual assessment of denoised maps,and statistical analysis of data distributions,as well as decomposition of uncertainties.Results show that the SC-VAE achieves superior noise suppression and better preservation of spatial structure compared to the standard VAE,as demonstrated by a significant reduction in the variogram nugget effect and an increased partial sill.The SC-VAE produces denoised maps with clearer anomaly delineation and more regularized data distributions,effectively mitigating outliers and reducing kurtosis.Additionally,it delivers improved interpolation accuracy and spatially explicit uncertainty estimates,facilitating more reliable and interpretable assessments of prediction confidence.The SC-VAE framework thus provides a robust,geostatistically informed solution for enhancing the quality and interpretability of geochemical data,with broad applicability in mineral exploration,environmental geochemistry,and other Earth Science domains.
基金supported by“Regional Innovation Strategy(RIS)”through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(MOE)(2021RIS-002)the Technology Development Program(RS-2023-00266141)funded by the Ministry of SMEs and Startups(MSS,Republic of Korea).
文摘Virtual Power Plants(VPPs)are integral to modern energy systems,providing stability and reliability in the face of the inherent complexities and fluctuations of solar power data.Traditional anomaly detection methodologies often need to adequately handle these fluctuations from solar radiation and ambient temperature variations.We introduce the Memory-Enhanced Autoencoder with Adversarial Training(MemAAE)model to overcome these limitations,designed explicitly for robust anomaly detection in VPP environments.The MemAAE model integrates three principal components:an LSTM-based autoencoder that effectively captures temporal dynamics to distinguish between normal and anomalous behaviors,an adversarial training module that enhances system resilience across diverse operational scenarios,and a prediction module that aids the autoencoder during the reconstruction process,thereby facilitating precise anomaly identification.Furthermore,MemAAE features a memory mechanism that stores critical pattern information,mitigating overfitting,alongside a dynamic threshold adjustment mechanism that adapts detection thresholds in response to evolving operational conditions.Our empirical evaluation of the MemAAE model using real-world solar power data shows that the model outperforms other comparative models on both datasets.On the Sopan-Finder dataset,MemAAE has an accuracy of 99.17%and an F1-score of 95.79%,while on the Sunalab Faro PV 2017 dataset,it has an accuracy of 97.67%and an F1-score of 93.27%.Significant performance advantages have been achieved on both datasets.These results show that MemAAE model is an effective method for real-time anomaly detection in virtual power plants(VPPs),which can enhance robustness and adaptability to inherent variables in solar power generation.
文摘We analyze the suitability of existing pre-trained transformer-based language models(PLMs)for abstractive text summarization on German technical healthcare texts.The study focuses on the multilingual capabilities of these models and their ability to perform the task of abstractive text summarization in the healthcare field.The research hypothesis was that large language models could perform high-quality abstractive text summarization on German technical healthcare texts,even if the model is not specifically trained in that language.Through experiments,the research questions explore the performance of transformer language models in dealing with complex syntax constructs,the difference in performance between models trained in English and German,and the impact of translating the source text to English before conducting the summarization.We conducted an evaluation of four PLMs(GPT-3,a translation-based approach also utilizing GPT-3,a German language Model,and a domain-specific bio-medical model approach).The evaluation considered the informativeness using 3 types of metrics based on Recall-Oriented Understudy for Gisting Evaluation(ROUGE)and the quality of results which is manually evaluated considering 5 aspects.The results show that text summarization models could be used in the German healthcare domain and that domain-independent language models achieved the best results.The study proves that text summarization models can simplify the search for pre-existing German knowledge in various domains.
文摘Drive-by techniques for bridge health monitoring have drawn increasing attention from researchers and practitioners,in the attempt to make bridge condition-based monitoring more cost-efficient.In this work,the authors propose a drive-by approach that takes advantage from bogie vertical accelerations to assess bridge health status.To do so,continuous wavelet transform is combined with multiple sparse autoencoders that allow for damage detection and localization across bridge span.According to authors’best knowledge,this is the first case in which an unsupervised technique,which relies on the use of sparse autoencoders,is used to localize damages.The bridge considered in this work is a Warren steel truss bridge,whose finite element model is referred to an actual structure,belonging to the Italian railway line.To investigate damage detection and localization performances,different operational variables are accounted for:train weight,forward speed and track irregularity evolution in time.Two configurations for the virtual measuring channels were investigated:as a result,better performances were obtained by exploiting the vertical accelerations of both the bogies of the leading coach instead of using only one single acceleration signal.
