Autoencoder Network Research Articles

SpATr: MoCap 3D human action recognition based on spiral auto-encoder and transformer network

Recent technological advancements have significantly expanded the potential of human action recognition through harnessing the power of 3D data. This data provides a richer understanding of actions, including depth information that enables more accurate analysis of spatial and temporal characteristics. In this context, We study the challenge of 3D human action recognition. Unlike prior methods, that rely on sampling 2D depth images, skeleton points, or point clouds, often leading to substantial memory requirements and the ability to handle only short sequences, we introduce a novel approach for 3D human action recognition, denoted as SpATr (Spiral Auto-encoder and Transformer Network), specifically designed for fixed-topology mesh sequences. The SpATr model disentangles space and time in the mesh sequences. A lightweight auto-encoder, based on spiral convolutions, is employed to extract spatial geometrical features from each 3D mesh. These convolutions are lightweight and specifically designed for fix-topology mesh data. Subsequently, a temporal transformer, based on self-attention, captures the temporal context within the feature sequence. The self-attention mechanism enables long-range dependencies capturing and parallel processing, enabling scalability for long sequences. The proposed method is evaluated on three prominent 3D human action datasets: Babel, MoVi, and BMLrub, from the Archive of Motion Capture As Surface Shapes (AMASS). Our results analysis demonstrates the competitive performance of our SpATr model in 3D human action recognition while maintaining efficient memory usage. The code and the training results are publicly available at https://github.com/h-bouzid/spatr.

Wind turbine generator early fault diagnosis using LSTM-based stacked denoising autoencoder network and stacking algorithm

ABSTRACT To reduce the significant economic losses caused by the fault deterioration of wind turbine generators, it is urgent to detect and diagnose the early faults of generators. The existing condition monitoring and fault diagnosis (CMFD) methods have disadvantages of less considering data temporal characteristic, acquiring early faults with difficulty, and having lower diagnostic accuracy. To address those limitations, a novel LSDAE-stacking CMFD method of generators was proposed. Specifically, a multivariate spatiotemporal condition monitoring model (LSDAE) was established by combining the LSTM and SDAE networks, which can detect generator early anomalies through real-time monitoring the reconstruction residual. Then, based on the stacking ensemble algorithm, a multi-classification fault diagnosis model (Stacking) was constructed to identify early fault types, which can integrate advantages of different base-classifiers to achieve a better diagnostic accuracy. Case studies on three actual generator failures were employed to validate the effectiveness and accuracy of the proposed LSDAE-stacking method. The results illustrated that, compared with conventional SDAE model, the proposed LSDAE model had higher reconstruction precision and superior early-fault-warning capacities. And compared with traditional algorithms such as SVM, RF, AdaBoost, GBDT and XGBoost, the constructed Stacking model can effectively identify the fault types of generators and had higher diagnostic accuracy.

Enhanced Regularized Ensemble Encoderdecoder Network for Accurate Brain Tumor Segmentation.

Segmenting tumors in MRI scans is a difficult and time-consuming task for radiologists. This is because tumors come in different shapes, sizes, and textures, making them hard to identify visually. This study proposes a new method called the enhanced regularized ensemble encoder-decoder network (EREEDN) for more accurate brain tumor segmentation. The EREEDN model first preprocesses the MRI data by normalizing the intensity levels. It then uses a series of autoencoder networks to segment the tumor. These autoencoder networks are trained using back-propagation and gradient descent. To prevent overfitting, the EREEDN model also uses L2 regularization and dropout mechanisms. The EREEDN model was evaluated on the BraTS 2020 dataset. It achieved high performance on various metrics, including accuracy, sensitivity, specificity, and dice coefficient score. The EREEDN model outperformed other methods on the BraTS 2020 dataset. The EREEDN model is a promising new method for brain tumor segmentation. It is more accurate and efficient than previous methods. Future studies will focus on improving the performance of the EREEDN model on complex tumors.

Artificial Intelligence in Fluorescence Lifetime Imaging Ophthalmoscopy (FLIO) Data Analysis-Toward Retinal Metabolic Diagnostics.

The purpose of this study was to investigate the possibility of implementing an artificial intelligence (AI) approach for the analysis of fluorescence lifetime imaging ophthalmoscopy (FLIO) data even with small data. FLIO data, including the fluorescence intensity and mean fluorescence lifetime (τm) of two spectral channels, as well as OCT-A data from 26 non-smokers and 28 smokers without systemic and ocular diseases were used. The analysis was performed with support vector machines (SVMs), a well-known AI method for small datasets, and compared with the results of convolutional neural networks (CNNs) and autoencoder networks. The SVM was the only tested AI method, which was able to distinguish τm between non-smokers and heavy smokers. The accuracy was about 80%. OCT-A data did not show significant differences. The feasibility and usefulness of the AI in analyzing FLIO and OCT-A data without any apparent retinal diseases were demonstrated. Although further studies with larger datasets are necessary to validate the results, the results greatly suggest that AI could be useful in analyzing FLIO-data even from healthy subjects without retinal disease and even with small datasets. AI-assisted FLIO is expected to greatly advance early retinal diagnosis.

Residual life prediction of multi-layer composite coatings based on vector quantised and attention-variational AutoEncoder network

ABSTRACT The quality of multi-layer composite coatings (MLCC) directly affects the reliability of equipment. Ultrasonic C-scan has advantages of non-destructive and high-precision. The accurate determination of residual life (RL) from C-scan images of coatings is hindered by the difficulty of feature extraction. To address this issue, a RL prediction method for MLCC based on the Vector Quantised and Attention-Variational AutoEncoder (VQA-VAE) network is proposed. First, by incorporating attention mechanism and residual module, the VQA-VAE network can perform unsupervised feature extraction on C-scan images effectively. Then, the features are fused by two-dimensional entropy weighting method to get the health index (HI), which can be used to establish the RL prediction model. To verify the effectiveness of this method, C-scan is performed on MLCC by ultrasonic microscope to obtain C-scan images, and RL of MLCC is predicted. The experiment results demonstrate that this method achieves higher accuracy in predicting RL of MLCC. In a word, this is a non-destructive, high-efficiency, high-precision, and large-scale measurement technique for MLCC.

Deep learning-based prediction of saturation magnetic flux density in Fe-based metallic glasses via transfer learning

Transfer learning (TL) is a widely recognized machine learning model that leverages domain knowledge from one or more predictive tasks to enhance performance in a related task. This study demonstrated the improved prediction performance of saturation magnetic flux density (Bs) of bulk metallic glasses (BMGs) with deep learning through the utilization of TL, which enabled the extension of compositional knowledge to overcome the limitations posed by scarce data. Specifically, TL was demonstrated by training a convolutional autoencoder network on full BMG datasets and transferring the acquired compositional knowledge to scarce iron (Fe)-based metallic glass datasets. Through this transfer, the broad compositional knowledge derived from the full BMG dataset is embedded in data representations, allowing its domain knowledge to be shared with smaller datasets. Here, a convolutional neural network was used as a predictor, with the prediction performance of Bs evaluated by the determination coefficient (R2 score) and root-mean-squared error (RMSE). The predictor, having received knowledge about the broad composition through TL, exhibited enhanced predictive performance of Bs on unseen BMG datasets while maintaining its prediction capability for Fe-based metallic glass alloys with R2 = 0.939 and RMSE = 0.084 T. This performance surpassed the performance of directly transferred knowledge on scarce data. These findings underscore the benefits of TL in predictive tasks with limited datasets and highlight the potential for applying this method to the development of new materials, where well-trained models can be repurposed for unlabeled material datasets with desired properties.

Deep learning assisted XRF spectra classification

EDXRF spectrometry is a well-established and often-used analytical technique in examining materials from which cultural heritage objects are made. The analytical results are traditionally subjected to additional multivariate analysis for archaeometry studies to reduce the initial data’s dimensionality based on informative features. Nowadays, artificial intelligence (AI) techniques are used more for this purpose. Different soft computing techniques are used to improve speed and accuracy. Choosing the most suitable AI method can increase the sustainability of the analytical process and postprocessing activities. An autoencoder neural network has been designed and used as a dimension reduction tool of initial 40×2048\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$40 \ imes 2048$$\\end{document} data collected in the raw EDXRF spectra, containing information about the selected points’ elemental composition on the canvas paintings’ surface. The autoencoder network design enables the best possible reconstruction of the original EDXRF spectrum and the most informative feature extraction, which has been used for dimension reduction. Such configuration allows for efficient classification algorithms and their performances. The autoencoder neural network approach is more sustainable, especially in processing time consumption and experts’ manual work.

Synthesis of Hybrid Data Consisting of Chest Radiographs and Tabular Clinical Records Using Dual Generative Models for COVID-19 Positive Cases.

To generate synthetic medical data incorporating image-tabular hybrid data by merging an image encoding/decoding model with a table-compatible generative model and assess their utility. We used 1342 cases from the Stony Brook University Covid-19-positive cases, comprising chest X-ray radiographs (CXRs) and tabular clinical data as a private dataset (pDS). We generated a synthetic dataset (sDS) through the following steps: (I) dimensionally reducing CXRs in the pDS using a pretrained encoder of the auto-encoding generative adversarial networks (αGAN) and integrating them with the correspondent tabular clinical data; (II) training the conditional tabular GAN (CTGAN) on this combined data to generate synthetic records, encompassing encoded image features and clinical data; and (III) reconstructing synthetic images from these encoded image features in the sDS using a pretrained decoder of the αGAN. The utility of sDS was assessed by the performance of the prediction models for patient outcomes (deceased or discharged). For the pDS test set, the area under the receiver operating characteristic (AUC) curve was calculated to compare the performance of prediction models trained separately with pDS, sDS, or a combination of both. We created an sDS comprising CXRs with a resolution of 256 × 256 pixels and tabular data containing 13 variables. The AUC for the outcome was 0.83 when the model was trained with the pDS, 0.74 with the sDS, and 0.87 when combining pDS and sDS for training. Our method is effective for generating synthetic records consisting of both images and tabular clinical data.

Generative approaches for solving tangram puzzles

The Tangram is a dissection puzzle composed of seven polygonal pieces that can form different patterns. Solving the Tangram is an irregular shape packing problem known to be NP-hard. This paper investigates the application of four deep-learning architectures—Convolutional Autoencoder, Variational Autoencoder, U-Net, and Generative Adversarial Network—specifically designed for solving Tangram puzzles. We explore the potential of these architectures in learning the complex spatial relationships inherent in Tangram configurations. Our experiments show that the Generative Adversarial Network competes well with other architectures and converges considerably faster. We further prove that traditional evaluation metrics based on pixel accuracy often fail in assessing the visual quality of the generated Tangram solutions. We introduce a loss function based on a Weighted Mean Absolute Error that prioritizes pixels representing inter-piece sections over those covered by individual pieces. Extending this loss function, we propose a novel evaluation metric as a more fitting measure for assessing Tangram solutions compared to traditional metrics. This investigation advances our understanding of the capabilities of artificial intelligence in complex geometrical problem domains.

Brain tumor classification from MRI scans: a framework of hybrid deep learning model with Bayesian optimization and quantum theory-based marine predator algorithm.

Brain tumor classification is one of the most difficult tasks for clinical diagnosis and treatment in medical image analysis. Any errors that occur throughout the brain tumor diagnosis process may result in a shorter human life span. Nevertheless, most currently used techniques ignore certain features that have particular significance and relevance to the classification problem in favor of extracting and choosing deep significance features. One important area of research is the deep learning-based categorization of brain tumors using brain magnetic resonance imaging (MRI). This paper proposes an automated deep learning model and an optimal information fusion framework for classifying brain tumor from MRI images. The dataset used in this work was imbalanced, a key challenge for training selected networks. This imbalance in the training dataset impacts the performance of deep learning models because it causes the classifier performance to become biased in favor of the majority class. We designed a sparse autoencoder network to generate new images that resolve the problem of imbalance. After that, two pretrained neural networks were modified and the hyperparameters were initialized using Bayesian optimization, which was later utilized for the training process. After that, deep features were extracted from the global average pooling layer. The extracted features contain few irrelevant information; therefore, we proposed an improved Quantum Theory-based Marine Predator Optimization algorithm (QTbMPA). The proposed QTbMPA selects both networks' best features and finally fuses using a serial-based approach. The fused feature set is passed to neural network classifiers for the final classification. The proposed framework tested on an augmented Figshare dataset and an improved accuracy of 99.80%, a sensitivity rate of 99.83%, a false negative rate of 17%, and a precision rate of 99.83% is obtained. Comparison and ablation study show the improvement in the accuracy of this work.

Research on a Feature Enhancement Extraction Method for Underwater Targets Based on Deep Autoencoder Networks

The low-frequency line spectrum of the radiated noise signals of hydroacoustic targets contains features describing the intrinsic properties of the target that make the target susceptible to exposure. In order to extract the line spectral features of underwater acoustic targets, a method combining image processing and a deep autoencoder network (DAE) is proposed in this paper to enhance the low-frequency weak line spectrum of underwater targets in an extremely low signal-to-noise ratio environment based on the measured data of large underwater vehicles. A Gauss–Bernoulli restricted Boltzmann machine (G–BRBM) for real-value signal processing was designed and programmed by introducing a greedy algorithm. On this basis, the encoding and decoding mechanism of the DAE network was used to eliminate interference from environmental noise. The weak line spectrum features were effectively enhanced and extracted under an extremely low signal-to-noise ratio of 10–300 Hz, after which the reconstruction results of the line spectrum features were obtained. Data from large underwater vehicles detected by far-field sonar arrays were processed and the results show that the method proposed in this paper was able to adaptively enhance the line spectrum in a data-driven manner. The DAE method was able to achieve more than double the extractable line spectral density in the frequency band of 10–300 Hz. Compared with the traditional feature enhancement extraction method, the DAE method has certain advantages for the extraction of weak line spectra.

Bearing fault diagnosis based on variational autoencoder and non-local block wide kernel convolutional neural network

At present, convolutional neural network (CNN) is widely applied to bearing fault diagnosis However, the diagnosis performance will descend under the strong noise condition in the real industrial environment. Therefore, a denoising method named non-local block wide kernel CNN (NLBWCNN) is proposed based on wide convolution kernel and non-local block. Additionally, the data in the mechanical fault state is less than that in the health state in actual industrial production, which leads to the data imbalance problem. However, the fault classifier based on CNN needs a large amount of balanced data to train. Otherwise, it will not be fully trained, and thus its generalization ability will be affected. As a result, a method called VAE-NLBWCNN (variational autoencoder and NLBWCNN) is proposed for diagnosing bearing faults. The method employs variational autoencoder balanced the fault data. And then, the NLBWCNN is utilized to denoise and classify the fault data. The proposed VAE-NLBWCNN method is validated on three bearing datasets. The comparative experiments demonstrate that the proposed method can effectively expand unbalanced data and achieve the best performance in various noise conditions.

Autoencoder in Autoencoder Networks.

Modeling complex correlations on multiview data is still challenging, especially for high-dimensional features with possible noise. To address this issue, we propose a novel unsupervised multiview representation learning (UMRL) algorithm, termed autoencoder in autoencoder networks (AE2-Nets). The proposed framework effectively encodes information from high-dimensional heterogeneous data into a compact and informative representation with the proposed bidirectional encoding strategy. Specifically, the proposed AE2-Nets conduct encoding in two directions: the inner-AE-networks extract view-specific intrinsic information (forward encoding), while the outer-AE-networks integrate this view-specific intrinsic information from different views into a latent representation (backward encoding). For the nested architecture, we further provide a probabilistic explanation and extension from hierarchical variational autoencoder. The forward-backward strategy flexibly addresses high-dimensional (noisy) features within each view and encodes complementarity across multiple views in a unified framework. Extensive results on benchmark datasets validate the advantages compared to the state-of-the-art algorithms.

Feature Extraction With Stacked Autoencoders for EEG Channel Reduction in Emotion Recognition.

Emotion recognition by electroencephalogram (EEG) signals is one of the complex methods because the extraction and recognition of the features hidden in the signal are sophisticated and require a significant number of EEG channels. Presenting a method for feature analysis and an algorithm for reducing the number of EEG channels fulfills the need for research in this field. Accordingly, this study investigates the possibility of utilizing deep learning to reduce the number of channels while maintaining the quality of the EEG signal. A stacked autoencoder network extracts optimal features for emotion classification in valence and arousal dimensions. Autoencoder networks can extract complex features to provide linear and non- linear features which are a good representative of the signal. The accuracy of a conventional emotion recognition classifier (support vector machine) using features extracted from SAEs was obtained at 75.7% for valence and 74.4% for arousal dimensions, respectively. Further analysis also illustrates that valence dimension detection with reduced EEG channels has a different composition of EEG channels compared to the arousal dimension. In addition, the number of channels is reduced from 32 to 12, which is an excellent development for designing a small-size EEG device by applying these optimal features.

Noise Reduction in Photoplethysmography Signals Using a Convolutional Denoising Autoencoder With Unconventional Training Scheme.

We propose an efficient approach based on a convolutional denoising autoencoder (CDA) network to reduce motion and noise artifacts (MNA) from corrupted atrial fibrillation (AF) and non-AF photoplethysmography (PPG) data segments so that an accurate PPG-signal-derived heart rate can be obtained. Our method's main innovation is the optimization of the CDA performance for both rhythms using more AF than non-AF data for training the AF-specific CDA model and vice versa for the non-AF CDA network. To evaluate this unconventional training scheme, our proposed network was trained and tested on 25-sec PPG data segments from 48 subjects from two different databases-the Pulsewatch dataset and Stanford University's publicly available PPG dataset. In total, our dataset contains 10,773 data segments: 7,001 segments for training and 3,772 independent segments from out-of-sample subjects for testing. Using real-life corrupted PPG segments, our approach significantly reduced the average heart rate root mean square error (RMSE) of the reconstructed PPG segments by 45.74% and 23% compared to the corrupted non-AF and AF data, respectively. Further, our approach exhibited lower RMSE, and higher sensitivity and PPV for detected peaks compared to the reconstructed data produced by the alternative methods. These results show the promise of our approach as a reliable denoising method, which should be used prior to AF detection algorithms for an accurate cardiac health monitoring involving wearable devices. PPG signals collected from wearables are vulnerable to MNA, which limits their use as a reliable measurement, particularly in uncontrolled real-life environments.

Anomaly Detection With Memory-Augmented Adversarial Autoencoder Networks for Industry 5.0

Anomaly Detection With Memory-Augmented Adversarial Autoencoder Networks for Industry 5.0

RPCA-AENet: Clutter Suppression and Simultaneous Stationary Scene and Moving Targets Imaging in the Presence of Motion Errors.

Clutter suppression and ground moving target imaging in synthetic aperture radar (SAR) system have been receiving increasing attention for both civilian and military applications. The problem of clutter suppression and ground moving target imaging in practical applications is much more challenging due to the motion error of the radar platform. In this article, we focus on the problems of clutter suppression and simultaneous stationary and moving target imaging in the presence of motion errors. Specifically, we propose a robust principal component analysis autoencoder network (RPCA-AENet) in a single-channel SAR system. In RPCA-AENet, the encoder transforms the SAR echo into imaging results of stationary scene and ground moving targets, and the decoder regenerates the SAR echo using the obtained imaging results. The encoder is designed by the unfolded robust principal component analysis (RPCA), while the decoder is formulated into two dense layers and one additional layer. Joint reconstruction loss, entropy loss, and measurement distance loss are utilized to guide the training of the RPCA-AENet. Notably, the algorithm operates in a totally self-supervised form and requires no other labeled SAR data. The methodology was tested on numerical SAR data. These tests show that the proposed architecture outperforms other state-of-the-art methods.

Health monitoring of an assembly-type structure using raw time-domain EMI signals and a novel hybrid deep learning algorithm

Deep learning models are widely used to extract features from data through supervised or unsupervised methods. However, when these two methods are used separately, their generalization ability is insufficient. In this study, a structural health monitoring technology based on a novel hybrid deep learning (HDL) model and time-domain electromechanical impedance (TEMI) has been proposed. The HDL is a highly efficient combination of deep auto-encoder (DAE) and two-dimensional convolutional neural network (2D CNN). The DAE is used for unsupervised reconstruction of the high-level features of 1D TEMI signals, and then, the features extracted by DAE are arranged and fed into a supervised 2D CNN that is designed to learn the optimal feature representations of the multi-label signals. This increases effectively the model’s learnability and generalizability. The robustness of the developed approach has been tested using TEMI signals that are collected from a scaled model of a rectangular pipe gallery subjected to transversal joint damage. Results indicate that the maximal identification errors of joint damage severity are 2.4% and 9.33% when the corresponding structural conditions are used and unused by the training process of HDL, respectively.

Relevance variable selection variational auto-encoder network for quality-related nonlinear process monitoring

Quality-related process monitoring is essential for revealing changes in product quality and ensuring industrial safety. Therefore, it is crucial to distinguish enough quality-related features within the data. To learn the nonlinear characteristics and obtain quality-related features in process data, a novel process monitoring method named relevance variable selection variational auto-encoder (RVS-VAE) is proposed. Firstly, to enhance the strong quality-related variables and learn the time correlation, the normalized original data is weighted by mutual information and augmented into data matrices. Secondly, the RVS strategy is proposed to select the most quality-related variables from the latent features. These quality-related features are further utilized to monitor the process. What is more, when a fault is detected, relevance variable relative contribution (RVRC) method is presented to locate the fault variables. Finally, performance of the proposed RVS-VAE method is evaluated through comparative experiments on tennessee eastman (TE) process.

AutoTransOP: translating omics signatures without orthologue requirements using deep learning

The development of therapeutics and vaccines for human diseases requires a systematic understanding of human biology. Although animal and in vitro culture models can elucidate some disease mechanisms, they typically fail to adequately recapitulate human biology as evidenced by the predominant likelihood of clinical trial failure. To address this problem, we developed AutoTransOP, a neural network autoencoder framework, to map omics profiles from designated species or cellular contexts into a global latent space, from which germane information for different contexts can be identified without the typically imposed requirement of matched orthologues. This approach was found in general to perform at least as well as current alternative methods in identifying animal/culture-specific molecular features predictive of other contexts—most importantly without requiring homology matching. For an especially challenging test case, we successfully applied our framework to a set of inter-species vaccine serology studies, where 1-to-1 mapping between human and non-human primate features does not exist.