Search

• Proposed a method (DEEP-EP) for prediction of Epigenetic Protein. • Features are extracted by DDE, DPC, and GAAC. • ERCNN is used for model training. • DEEP-EP secured the best performance. Epigenetic proteins (EP) play a role in the progression of a wide range of diseases, including autoimmune disorders, neurological disorders, and cancer. Recognizing their different functions has prompted researchers to investigate them as potential therapeutic targets and pharmacological targets. This paper proposes a novel deep learning-based model that accurately predicts EP. This study introduces a novel deep learning-based model that accurately predicts EP. Our approach entails generating two distinct datasets for training and evaluating the model. We then use three distinct strategies to transform protein sequences to numerical representations: Dipeptide Deviation from Expected Mean (DDE), Dipeptide Composition (DPC), and Group Amino Acid (GAAC). Following that, we train and compare the performance of four advanced deep learning models algorithms: Ensemble Residual Convolutional Neural Network (ERCNN), Generative Adversarial Network (GAN), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU). The DDE encoding combined with the ERCNN model demonstrates the best performance on both datasets. This study demonstrates deep learning's potential for precisely predicting EP, which can considerably accelerate research and streamline drug discovery efforts. This analytical method has the potential to find new therapeutic targets and advance our understanding of EP activities in disease. [ABSTRACT FROM AUTHOR]

Currently, deep learning has achieved remarkable success in estimating plant disease from unmanned aerial system (UAS) images. However, two critical challenges remain unexplored: spatiotemporal variations in disease symptoms and the domain shift between source and target datasets. To overcome these challenges, this paper proposes an approach that incorporates temporal aspects of disease progression using time series analysis. Spatiotemporal information is integrated by combining convolutional neural networks and bidirectional long-short term memory (CNN-BiLSTM) to classify the disease into five severity levels. Various feature extraction methods, including both handcrafted and CNN-based feature extractors, are evaluated. Furthermore, to tackle the problem of domain shift, a feature-level domain adaptation method is proposed. This method aims to learn transferable feature representations that remain consistent despite variations between source and target datasets. This approach enhances the spatiotemporal transferability of the CNN-BiLSTM model, enabling the effective utilisation of historical datasets. The study demonstrates that the CNN-BiLSTM model outperforms traditional time-independent machine-learning methods that rely on handcrafted features. Specifically, the Resnet101-BiLSTM model achieves the highest overall classification accuracy of 89.7% among all tested models evaluated on a one-year dataset. Moreover, it shows superior generalisation with 72.7% accuracy for cross-spatiotemporal disease severity classification using domain adaptation, as demonstrated through two-year experiments. By reducing the domain shift of source and target datasets and harnessing time series high-resolution images obtained throughout the crop growing season, this hybrid approach has substantial potential to advance the assessment of crop disease severity in field conditions. • Compared different CNN-based feature extractors for FOV4 symptom detection in UAS images. • Developed a BiLSTM model for temporal analysis in UAS datasets. • Use of domain adaptation enhanced model generalization. • CNN-BiLSTM with CORAL achieved 72.7% accuracy, proving its effectiveness in diverse real-world applications. [ABSTRACT FROM AUTHOR]

• We introduce a hexagonal convolution neural network for images with hexagonal pixels. • We applied hexCNN to segment spatially resolved transcriptomics data. • HexCNN achieved greater segmentation accuracies than state-of-the-art GNN models. • We noticed considerable dropout events and batch effects hindering further progress. Spatial transcriptomics is a rapidly evolving field that enables researchers to capture comprehensive molecular profiles while preserving information about the physical locations. One major challenge in this research area involves the identification of spatial domains, which are distinct regions characterized by unique gene expression patterns. However, current unsupervised methods have struggled to perform well in this regard due to the presence of high levels of noise and dropout events in spatial transcriptomic profiles. In this paper, we propose a novel hexagonal Convolutional Neural Network (hexCNN) for hexagonal image segmentation on spatially resolved transcriptomics. To address the problem of noise and dropout occurrences within spatial transcriptomics data, we first extend an unsupervised algorithm to a supervised learning method that can identify useful features and reduce noise hindrance. Then, inspired by the classical convolution in convolutional neural networks (CNNs), we designed a regular hexagonal convolution to compensate for the missing gene expression patterns from adjacent spots. We evaluated the performance of hexCNN by applying it to the DLPFC dataset. The results show that hexCNN achieves a classification accuracy of 86.8% and an average Rand index (ARI) of 77.1% (1.4% and 2.5% higher than those of GNNs). The results also demonstrate that hexCNN is capable of removing the noise caused by batch effect while preserving the biological signal differences. [ABSTRACT FROM AUTHOR]

Although code generation for Convolution Neural Network (CNN) models has been extensively studied, performing efficient data slicing and parallelization for highly-constrained Multicore Neural Processor Units (NPUs) is still a challenging problem. Given the size of convolutions' input/output tensors and the small footprint of NPU on-chip memories, minimizing memory transactions while maximizing parallelism and MAC utilization are central to any effective solution. This paper proposes a TensorFlow XLA/LLVM compiler optimization pass for Multicore NPUs, called Tensor Slicing Optimization (TSO), which: (a) maximizes convolution parallelism and memory usage across NPU cores; and (b) reduces data transfers between host and NPU on-chip memories by using DRAM memory burst time estimates to guide tensor slicing. To evaluate the proposed approach, a set of experiments was performed using the NeuroMorphic Processor (NMP), a multicore NPU containing 32 RISC-V cores extended with novel CNN instructions. Experimental results show that TSO is capable of identifying the best tensor slicing that minimizes execution time for a set of CNN models. Speed-ups of up to 21.7% result when comparing the TSO burst-based technique to a no-burst data slicing approach. To validate the generality of the TSO approach, the algorithm was also ported to the Glow Machine Learning framework. The performance of the models were measured on both Glow and TensorFlow XLA/LLVM compilers, revealing similar results. • A burst-based cost model for Tensor Slicing Optimization. • Reducing data transfer between host and NPU on-chip memories. • Maximizing Convolution parallelism and memory reuse. [ABSTRACT FROM AUTHOR]

Chlorophyll-a (Chl-a) concentration plays a crucial role in monitoring marine phytoplankton, holding significant implications for marine ecosystems and human livelihoods. Remote sensing serves as a valuable method for directly estimating Chl-a concentration and facilitating marine monitoring activities. In order to accurately estimate Chl-a concentration and effectively monitor the marine ecological environment of the Taiwan Strait, this paper introduces a novel neural network model called Convolutional Neural Network with Dual Attention Mechanism Optimization (CNN-CBAM). Additionally, the spectral bands derived from MODIS 500 m imagery are subjected to various processing techniques to generate four distinct sets of feature data. These datasets are subsequently merged with in-situ measurements of Chl-a concentration obtained from buoy stations. Given the limited availability of buoy stations, the temporal continuity of data is leveraged to overcome the spatial distribution constraints. This enables the estimation of Chl-a concentration at various time intervals within the research area of the Taiwan Strait. To evaluate the performance of the model, metrics such as root mean square error (RMSE), mean absolute percentage error (MAPE), and R2 are employed. The results demonstrate that the CNN-CBAM model, utilizing MODIS temporal data, effectively retrieves Chl-a concentration in the research area of the Taiwan Strait with high accuracy. For the CNN-CBAM model, the RMSE is 0.34 μg L−1, MAPE is 22%, and R2 is 0.86. In comparison, the traditional empirical model using the Band Ratio (BR) model has an RMSE of 0.61 μg L−1, MAPE of 57%, and R2 of 0.31. The results indicate that the CNN-CBAM model has improved R2 by 0.55 compared to the traditional BR model. Additionally, the RMSE and MAPE have decreased by 0.27 μg L−1 and 35%, respectively. The superiority of the CNN-CBAM model over the traditional BR model is evident, showcasing a substantial enhancement in the accuracy of Chl-a concentration retrieval. Given the challenges posed by the intricate and ever-changing marine environment, particularly the scarcity of in-situ buoy stations, this study offers a practical solution for monitoring Chl-a concentration. It establishes a framework for leveraging satellite-based multispectral sensors to enable large-scale and multi-temporal remote sensing retrieval of Chl-a concentration in the research area of the Taiwan Strait. • The CNN-CBAM model demonstrates exceptional performance in estimating Chl-a concentration. • The CNN-CBAM model excels in mining non-linear data compared to traditional empirical models. • CNN-CBAM outperforms other machine learning models in extracting high-dimensional feature data. • The incorporation of the CBAM module enhances the performance of the CNN model. [ABSTRACT FROM AUTHOR]

Accurate prediction of carbon price is of great significance to national energy security and climate environment policies. This paper comes up with a new forecasting model variational mode decomposition, convolutional neural network, bidirectional long short-term memory, and multi-layer perceptron (VMD–CNN–BILSTM-MLP) to predict EUA carbon futures prices in two periods of five years before and after the introduction of emission reduction policies. The parameters of the VMD model are determined by genetic algorithm (GA) firstly, carbon futures prices are broken down into subsequences of different frequencies using the model. The MLP model is then applied to predict the highest frequency sequence. The CNN-BILSTM model is applied to predict other subsequences later. Finally, the predicted values of each subsequence are linearly added to obtain the final result of the entire model. The prediction effect of the model is mainly tested by root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), coefficient of determination (R2) and the modification of Diebold-Mariano test (MDM). In both periods, the proposed model predicts better than the other models, and the prediction effect of carbon futures price in the first five years is a little better than that in the second five years. In general, the experiment of predicting carbon futures prices in two different periods, the experiment of changing the proportion of data set and the experiment of predicting the whole sample all prove that the mixed model proposed in this paper has good prediction effect. • Carbon futures prices are predicted by a hybrid model VMD–CNN–BILSTM-MLP. • Genetic algorithm is used to optimize variational mode decomposition. • The hierarchical prediction based on subsequence features is better. • The model has great prediction effect before and after emission reduction policies. [ABSTRACT FROM AUTHOR]

• We propose a federated deep learning-based IDS for mitigating cyberattacks. • We investigate the use of three deep learning classifiers, DNN, CNN, and RNN. • The performance of each classifier is evaluated using three recent real datasets. • The results show that the FELIDS system outperforms the centralized versions. • The proposed FELIDS model achieves the highest accuracy in detecting attacks. In this paper, we propose a federated learning-based intrusion detection system, named FELIDS, for securing agricultural-IoT infrastructures. Specifically, the FELIDS system protects data privacy through local learning, where devices benefit from the knowledge of their peers by sharing only updates from their model with an aggregation server that produces an improved detection model. In order to prevent Agricultural IoTs attacks, the FELIDS system employs three deep learning classifiers, namely, deep neural networks, convolutional neural networks, and recurrent neural networks. We study the performance of the proposed IDS on three different sources, including, CSE-CIC-IDS2018, MQTTset, and InSDN. The results demonstrate that the FELIDS system outperforms the classic/centralized versions of machine learning (non-federated learning) in protecting the privacy of IoT devices data and achieves the highest accuracy in detecting attacks. [ABSTRACT FROM AUTHOR]

Due to issues such as limited variability in monitoring data across different tool wear conditions and interference during the machining process, data-driven monitoring models are susceptible to misclassification. Therefore, this paper proposes a pioneering approach that takes into account the tool wear law and the characteristic distribution of tool wear monitoring data. Specifically, the paper proposes an automatic feature extraction and tool condition monitoring method based on Siamese Long Short-term Memory Networks (SLSTMs) to transform the original data distribution, enhance the differentiation of different tool wear condition monitoring data, and achieve accurate prediction of tool wear condition. Additionally, a hybrid data and mechanism-driven tool wear monitoring method is proposed that takes advantage of the fact that tool wear is continuous and irreversible to enable reliable adjustment of the monitoring results of the data-driven model without human intervention. The study conducted milling experiments on a three-axis vertical machining center, using TA2 titanium alloy as the workpiece material. Vibration signals from the spindle in three directions were collected as input to the network, with tool wear state labeled as "0″ or "1″ as the output. Classical machine learning algorithms such as Decision Trees, K-Nearest Neighbors (KNN) and Support Vector Machines (SVM), as well as classical deep learning algorithms such as Convolutional Neural Networks (CNN), Sparse Stacked Autoencoders (SSAE) and Long Short-Term Memory Neural Networks (LSTM), were used to construct and compare feature extraction and state monitoring models. The proposed model based on Siamese Long Short-Term Memory Neural Networks (SLSTMs) achieved testing average accuracy of 98.2% (σ2 = 0.44), outperforming typical deep learning algorithms such as CNN, LSTM and SSAE as well as traditional machine learning algorithms such as Decision Trees, KNN and SVM in terms of accuracy and robustness. Moreover, the proposed data and mechanism hybrid-driven tool wear monitoring method can effectively improve the monitoring accuracy and robustness of data-driven models such as CNN and SSAE. The monitoring accuracy of CNN and SSAE increased from 95.9% to 97.6% and 91.0% to 94.3%, respectively. [ABSTRACT FROM AUTHOR]

With the rapidly growing importance of biological research, non-coding RNAs (ncRNA) attract more attention in biology and bioinformatics. They play vital roles in biological processes such as transcription and translation. Classification of ncRNAs is essential to our understanding of disease mechanisms and treatment design. Many approaches to ncRNA classification have been developed, several of which use machine learning and deep learning. In this paper, we construct a novel deep learning-based architecture, ncRDense, to effectively classify and distinguish ncRNA families. In a comparative study, our model produces comparable results with existing state-of-the-art methods. Finally, we built a freely accessible web server for the ncRDense tool, which is available at http://nsclbio.jbnu.ac.kr/tools/ncRDense/. • Computational predictor is developed for prediction of Non-coding RNA family using deep learning. • Novel dense convolution neural network is proposed. • Achieved promising outcomes than existing methods. • A webserver tool is developed and made available for other researchers. [ABSTRACT FROM AUTHOR]

Auto-scaling systems can dynamically scale the required resources for cloud-based services at runtime. This is an effective mechanism, enabling services to adapt to environmental changes. These systems establish the foundation for achieving elasticity in the modern cloud computing paradigm. Given the dynamic and uncertain nature of the shared cloud infrastructure, cloud auto-scaling systems are one of the most complex and sophisticated created artifacts, aiming to achieve self-aware, self-adaptive, and dependable runtime scaling. To find an effective solution to this problem, an accurate prediction of the required amount of workload as well as the system metrics for future time periods are needed. Various solutions have already been proposed to tackle this problem. Many solutions make use of machine learning, statistical, and ensemble methods. In this paper, we view the auto-scaling problem as a sequence model and apply the convolutional neural networks to predict the future workload of cloud services. Also, by using neural networks, we obtain a mapping between the predicted workload as well as the real-time and future amounts of the required resources. We have also proposed a decision-making mechanism that takes into account different and sometimes conflicting user criteria resulting in the best-compromised decision. To this aim, we have used TOPSIS as a multi-criteria decision-making method for the decision-making component. In the evaluation section, we have examined the amount of prediction error, the amount of service level agreement violations, as well as the amount of resources' under-utilization. Evaluations demonstrate that the proposed approach for predicting the workload shows a 4 percent improvement over the existing approaches. • Auto-scaling systems can dynamically scale the required resources for cloud services. • This problem is viewed as a sequence model. • The convolutional neural networks are applied to predict the workload of cloud services. • A decision-making mechanism taking into account different user criteria is proposed. • Evaluations demonstrate a 4 percent improvement over the existing approaches. [ABSTRACT FROM AUTHOR]

The Allen-Cahn equation is a well-known stiff semilinear parabolic equation used to describe the process of phase separation and transition in phase field modeling of multi-component physical systems, while the conservative Allen-Cahn equation is a modified version of the classic Allen-Cahn equation that can additionally conserve the mass. As neural networks and deep learning techniques have achieved significant successes in recent years in scientific and engineering applications, there has been growing interest in developing deep learning algorithms for numerical solutions of partial differential equations. In this paper, we propose a novel deep learning method for predicting the dynamics of the classic and conservative Allen-Cahn equations. Specifically, we design two special convolutional neural network models, one for each of the two equations, to learn the fully-discrete operators between two adjacent time steps. The loss functions of the two models are defined using the residual of the fully-discrete systems, which result from applying the central finite difference discretization in space and the Crank–Nicolson approximation in time. This approach enables us to train the models without requiring any ground-truth data. Moreover, we introduce an effective training strategy that automatically generates useful samples along the time evolution to facilitate training of the models. Finally, we conduct extensive experiments in two and three dimensions to demonstrate outstanding performance of our proposed method, including its dynamics prediction and generalization ability under different scenarios. • We propose two novel convolutional neural networks to learn the fully-discrete operators of the classic and conservative Allen-Chan equations. • The loss functions are defined using the residual of the fully-discrete systems, which enables us to train the models without ground-truth data. • We introduce a novel training strategy that automatically generates samples along the time to facilitate effective training of the two models. • Extensive experiments demonstrate outstanding performance of the proposed models for dynamics prediction under different scenarios. [ABSTRACT FROM AUTHOR]

• Brain decoding using a dense individual fMRI dataset (∼7 h fMRI data per subject). • Challenging benchmark: decoding 21 conditions from single fMRI brain volumes. • The best individual model reaches 58–67 % accuracy, close to state-of-the-art group decoding (76 %). • Individual decoding models generalized poorly to other subjects. Brain decoding aims to infer cognitive states from patterns of brain activity. Substantial inter-individual variations in functional brain organization challenge accurate decoding performed at the group level. In this paper, we tested whether accurate brain decoding models can be trained entirely at the individual level. We trained several classifiers on a dense individual functional magnetic resonance imaging (fMRI) dataset for which six participants completed the entire Human Connectome Project (HCP) task battery >13 times over ten separate fMRI sessions. We evaluated nine decoding methods, from Support Vector Machines (SVM) and Multi-Layer Perceptron (MLP) to Graph Convolutional Neural Networks (GCN). All decoders were trained to classify single fMRI volumes into 21 experimental conditions simultaneously, using ∼7 h of fMRI data per participant. The best prediction accuracies were achieved with GCN and MLP models, whose performance (57–67 % accuracy) approached state-of-the-art accuracy (76 %) with models trained at the group level on >1 K hours of data from the original HCP sample. Our SVM model also performed very well (54–62 % accuracy). Feature importance maps derived from MLP —our best-performing model— revealed informative features in regions relevant to particular cognitive domains, notably in the motor cortex. We also observed that inter-subject classification achieved substantially lower accuracy than subject-specific models, indicating that our decoders learned individual-specific features. This work demonstrates that densely-sampled neuroimaging datasets can be used to train accurate brain decoding models at the individual level. We expect this work to become a useful benchmark for techniques that improve model generalization across multiple subjects and acquisition conditions. [Display omitted] [ABSTRACT FROM AUTHOR]

The vibration signal analysis-based structural health monitoring (SHM) system has been widely used for bridge condition identification. With the development of artificial intelligence (AI), machine learning (ML) has made numerous breakthroughs in the detection of civil engineering structures. However, conventional data-driven ML methods heavily rely on prior knowledge for their performance. This paper proposes a deep learning (DL) model 1DCNN-BiLSTM for detecting small local structural changes of reinforced concrete (RC) beams, which applies the structure of the Inception module in GoogLeNet to one-dimensional convolution neural networks (1DCNNs) for feature extraction at different scales, and combines the advantages of bidirectional long short-term memory (BiLSTM) modules for processing long time-series data. Firstly, a three-dimensional numerical model of the RC beam was established using finite element software, and the accuracy of the numerical model was verified by comparing it with the test results. Based on the numerical model, tests on RC beams impacted by falling hammers were carried out, and the acceleration signals of the beam in different states were collected as a dataset. The proposed DL model can automatically extract the spatial and temporal domain features in the signals, accurately identify the location of small bridge local variations, the accuracy in the test set is 98.8% compared to 92.6% for a traditional ML model, and has better noise immunity performance and robustness to the missing data than the traditional ML model. Finally, the internal inference process of the model was explored and visualized, illustrating that the model has adaptive learning capabilities. [ABSTRACT FROM AUTHOR]

• Classified vibration signals to detect leakages in real-life pipe networks. • Implemented automated detection using accelerometers and machine learning methods. • Compared applied models based on multiple criteria against SqueezeNet. In addition to the issues of water quality and economic losses, leakages in pipe networks are a potential danger to public safety. Early detection of water leaks in pipe networks is crucial for implementing countermeasures to reduce structural damage and water losses. The method of data collection based on sensors installed in pipeline networks has gained a lot of attention due to its potential for application in real-time monitoring systems for leak detection. However, many models developed on data collected from simulations, engineered tests or field experiments, and few have been validated using real network data. To make further investigations in real water distribution systems (WDSs), machine learning (ML)-based leak detection models were developed and validated using signal data collected by piezoelectric accelerometers installed in WDSs over several cities of China. Nonlinear features describing the non-stationary characteristics contained in leak signals are extracted from the signals to build feature sets as inputs to the common classifiers, i.e., support vector machine (SVM), decision tree (DT), and K-nearest neighbor (KNN). In addition, the spectrograms obtained by the short-time Fourier transform processing are used for the convolutional neural network SqueezeNet. The effects of different features, different dataset compositions, and different models for leak detection are compared and analyzed. SqueezeNet has the best performance with 95.15% accuracy in identifying leaks, and KNN is the best of the three classifiers with superior sensitivity and 88.17% accuracy in identification. This paper demonstrates the effectiveness and practicability of using machine learning models to identify leaks in real pipe networks. The application of machine learning models can perform effective leak detection with minimal human intervention, thus providing a meaningful contribution to the study of leak detection in real WDSs. The proposed model aims to identify leakages under operational conditions in real-life pipe networks with higher efficiency and reduced manual involvement. This paper demonstrates the applicability of accelerometers based on microelectromechanical systems for leak detection and establishes real network-based machine learning models; thereby, providing meaningful contribution to the literature regarding leak detection in a real WDS. [ABSTRACT FROM AUTHOR]

Corrosion morphology prediction for civil infrastructure using physics-based simulation is computationally challenging due to the coupled multi-physics simulations involved in long-term corrosion prognostics. Machine learning (ML)-based surrogate modeling provides a promising way of overcoming this challenge. This paper presents a physics-constrained ML method for surrogate modeling of a high-fidelity multi-physics pitting corrosion simulation, which is solved using a phase-field method. The proposed surrogate model consists of a convolutional variational autoencoder to reduce the dimension of pitting corrosion shape images and a Bayesian multi-layer perceptron network (also known as the Bayesian Latent Space Time Evolution Network) to model the evolution of the corrosion pit morphology over time. To account for the fact that corrosion damage without repair is an irreversible process, a physics constraint is added to the surrogate model to ensure that the corrosion rate is strictly negative. The proposed physics-constrained surrogate modeling method is compared with a purely data-driven surrogate modeling method which integrates a convolutional neural network autoencoder with a Gaussian process regression (GPR) model-based nonlinear autoregressive exogenous model. The results show that the proposed physics-constrained surrogate model is significantly more accurate than the purely data-driven GPR-based surrogate model. [ABSTRACT FROM AUTHOR]

This paper presents a methodology for observing and analyzing marine ecosystems using images gathered from autonomous marine vehicles. Visual data is composed in photo-mosaics and classified using machine learning algorithms. The approach expands existing solutions, enabling extended monitoring in time, space, and depth. Imagery was collected during a field campaign in the Spanish marine and terrestrial protected area of Cabrera, Balearic Islands, colonized by the endemic seagrass species Posidonia oceanica (Po). The operations were performed using three distinct platforms, an Autonomous Underwater Vehicle (AUV), an Autonomous Surface Vehicle (ASV) and a Lagrangian Drifter (LD). Results are compared to prior habitat maps to assess seagrass meadow distribution. The proposed solution can be scaled and adapted to other locations and species, considering limitations in data storage and battery endurance. [Display omitted] • Seafloor habitat mapping with autonomous robots, filling gaps missed by divers. • Assessment of marine habitats using convolutional neural networks. • Methods apply to georeferenced imaging (e.g. drones, satellites) beyond subsea mapping. • Field campaign data from a marine area compared to previous habitat map data. [ABSTRACT FROM AUTHOR]

Predictive maintenance, as one of the core components of Industry 4.0, takes a proactive approach to maintain machines and systems in good order to keep downtime to a minimum and the airline maintenance industry is not an exception to this. To achieve this goal, practices in Structural Health Monitoring (SHM) complement the existing Non-Destructive-Testing (NDT) have been established in the last decades. Recently, the increasing computational capability such as utilization of a graphical processing unit (GPU) in combination with advanced machine learning techniques such as deep learning has been one of the main drivers in the advancement of predictive analytics in condition monitoring. In our previous work, we proposed a novel approach using deep learning for guided wave based structural health called DeepSHM. As a study case, we treated an ultrasonic signal from guided Lamb wave SHM with a convolutional neural network (CNN). In that work, we only considered a single central frequency excitation. This led to a single governing wavelength which is normally good for the detection of a single damage size. In classical signal processing, applying a broader excitation frequency poses an analysis and interpretation nightmare because it contains more complex information and thus is difficult to understand. This problem can be overcome with deep learning; however, it creates another problem: while deep learning typically results in a more accurate result prediction, it is specifically made for solving only certain types of tasks. While many papers have already introduced deep learning for diagnostics, many of these works are only proposing novel predictive techniques, however the mathematical formalization is lacking, and we are not informed about why we should treat acoustic signal with deep learning. So, the basis of 'explainable AI' for SHM and NDT is currently lacking. For this reason, in this paper, we would like to extend our previous work into a more generalized. Rather than focusing on a novel technique, we propose a plausible theoretical perspective inspired from neuroscience for signal representation of deep learning framework to model machine perception in structural health monitoring (SHM), especially because SHM typically involves multiple sensory input from different sensing locations. To do this, we created a set of artificial data from a finite element model (FEM) and represented DeepSHM in two different ways: 1). Perpetual representation of observation and 2). Hierarchical structure of entities that is decomposable in a smaller sub-entity. Consequently, we assume two plausible models for DeepSHM: 1). Either it behaves as a single deciding actor since the observation is regarded as perpetual, and 2). Or it acts as a multiple actor with independent outputs since multiple sensors can form different output probabilities. These artificial data were split into several different input representations, classified into several damage scenarios and then trained with commonly used deep learning training parameters. We compare the performance metrics of each perception model to describe the training behavior of both representations. [ABSTRACT FROM AUTHOR]

• A new method to predict and monitor concrete strength development. • The method adopts the state-of-the-art embedded smart aggregates, electromechanical impedance technique, and machine learning. • Conductance resonant frequency (CRF) and conductance resonant peak (CRP) are good concrete strength indicators. • The combined model with linear regression and convolutional neural network, has achieved the highest prediction accuracy. It is crucial to monitor the real-time strength development of early-age concrete, which is related to the structure safety and construction efficiency of concrete. This paper proposed a new method to predict and monitor concrete strength development. The basic idea is using embedded smart aggregates (SAs), combining the electromechanical impedance (EMI) technology and machine learning to monitor the strength development of early-age concrete. An EMI model of the embedded SA was firstly established and the relationship between the conductance resonant frequency (CRF) and conductance resonant peak (CRP) of the SA and the development of concrete strength was investigated. The conductance signatures were continuously monitored and the compressive strength of the specimens at different ages was tested. Hence, the EMI model results that the CRF decreases but CRP increases as the compressive strength of concrete increases were verified by analyzing the collected data. The four different models of a linear regression model (LR), a convolutional neural network (CNN), a hybrid (LR-CNN) model, and an empirical equation were established. The conductance of SAs as the developed models inputs, the compressive strength of specimens as outputs. Finally, these four models were compared with each other. Meanwhile, the performance indicators of the LR-CNN model and other hybrid models were compared. The results show that the LR-CNN model has excellent performance and successfully achieves quantitative prediction of concrete strength development. The proposed method for evaluating and predicting the compressive strength of concrete is simple, accurate, quantitative and reliable, and has promising application. [ABSTRACT FROM AUTHOR]

• GANs discretize continuous speech data into reduced representations that resemble rule- like generalizations. • We can actively observe the network's progress in learning by keeping latent space constant across different training steps. • GANs provide useful information for computational models of speech even if trained on a comparatively small dataset. • Local processes are easier to learn than non-local processes, which matches both behavioral data in human subjects and typology in the world's languages. • Rule-like generalizations in the training data are represented as an interaction between variables in the network's latent space. This paper argues that training Generative Adversarial Networks (GANs) on local and non-local dependencies in speech data offers insights into how deep neural networks discretize continuous data and how symbolic-like rule-based morphophonological processes emerge in a deep convolutional architecture. Acquisition of speech has recently been modeled as a dependency between latent space and data generated by GANs in Beguš (2020b), who models learning of a simple local allophonic distribution. We extend this approach to test learning of local and non-local phonological processes that include approximations of morphological processes. We further parallel outputs of the model to results of a behavioral experiment where human subjects are trained on the data used for training the GAN network. Four main conclusions emerge: (i) the networks provide useful information for computational models of speech acquisition even if trained on a comparatively small dataset of an artificial grammar learning experiment; (ii) local processes are easier to learn than non-local processes, which matches both behavioral data in human subjects and typology in the world's languages. This paper also proposes (iii) how we can actively observe the network's progress in learning and explore the effect of training steps on learning representations by keeping latent space constant across different training steps. Finally, this paper shows that (iv) the network learns to encode the presence of a prefix with a single latent variable; by interpolating this variable, we can actively observe the operation of a non-local phonological process. The proposed technique for retrieving learning representations has general implications for our understanding of how GANs discretize continuous speech data and suggests that rule-like generalizations in the training data are represented as an interaction between variables in the network's latent space. [ABSTRACT FROM AUTHOR]

• Machine learning model classifies acute vs chronic multiple sclerosis (MS) lesions. • Used single timepoint non-contrast conventional T1- and T2-weighted MRI only. • Identified 16 textural radiomic features discriminative of MS lesion subtype. • Achieved 74% balanced accuracy with 71–76% sensitivity and 73–78% specificity. Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease characterized by the appearance of focal lesions across the central nervous system. The discrimination of acute from chronic MS lesions may yield novel biomarkers of inflammatory disease activity which may support patient management in the clinical setting and provide endpoints in clinical trials. On a single timepoint and in the absence of a prior reference scan, existing methods for acute lesion detection rely on the segmentation of hyperintense foci on post-gadolinium T1-weighted magnetic resonance imaging (MRI), which may underestimate recent acute lesion activity. In this paper, we aim to improve the sensitivity of acute MS lesion detection in the single-timepoint setting, by developing a novel machine learning approach for the automatic detection of acute MS lesions, using single-timepoint conventional non-contrast T1- and T2-weighted brain MRI. The MRI input data are supplemented via the use of a convolutional neural network generating "lesion-free" reconstructions from original "lesion-present" scans using image inpainting. A multi-objective statistical ranking module evaluates the relevance of textural radiomic features from the core and periphery of lesion sites, compared within "lesion-free" versus "lesion-present" image pairs. Then, an ensemble classifier is optimized through a recursive loop seeking consensus both in the feature space (via a greedy feature-pruning approach) and in the classifier space (via model selection repeated after each pruning operation). This leads to the identification of a compact textural signature characterizing lesion phenotype. On the patch-level task of acute versus chronic MS lesion classification, our method achieves a balanced accuracy in the range of 74.3–74.6% on fully external validation cohorts. [ABSTRACT FROM AUTHOR]

Cloud identification on Mars is an important tool for climatology studies, making it possible to analyze the distribution, patterns and variability of clouds both spatially and temporally. Traditionally, cloud data on Mars has been extracted through manual or semi-automated processes which can be time consuming, and currently there is limited spatial and temporal cloud data coverage. In this paper we demonstrate the successful use of convolutional neural networks (CNNs) to extract cloud masks from Mars Daily Global Maps (MDGMs) composed from the Mars Color Imager (MARCI) on the Mars Reconnaissance Orbiter (MRO). The fully automated model reports an F 1 score of 0.7925 compared to the prior semi-automatic technique, and in some occasions the model performs better at extracting the full extent of the cloud. We also introduce several image pre- and post-processing techniques to improve the model's performance and usability. The model is configured to provide cloud masks at 0.1 ° longitude by 0.1 ° latitude resolution. It also automatically bounds the MDGM by northern and southern polar extents depending on solar longitude. The results suggest that our deep learning model is a useful tool to automatically and quickly extract Martian water ice cloud masks and make it possible to generate cloud mask data across the complete set of MDGMs and future ones. The model and related techniques also have potential extensions to Martian dust storm identification. Our code, model, and data are publicly available. • A Convolutional Neural Network is employed to identify martian water ice clouds. • This fully automated technique identifies clouds in Mars Daily Global Maps. • The model reaches >98% accuracy in cloud identification. • Over 5 Mars years are analyzed and publicly available online. [ABSTRACT FROM AUTHOR]

• This paper presents a deep learning network that inspects ultrasonic signals for defect detection. • The network is an adaptation of WaveNet, hence is based on causal dilated convolutional neural networks and residual blocks. • The network is shown to outperform both a widely used conventional analysis method and some competing deep learning algorithms. • The validation was performed on two datasets, one generated via finite element simulations on a steel plate, the other by actual experiments on a composite plate. • Significant improvements over the conventional method are obtained especially when testing occurs at temperatures well outside the range in the training set of signals. This paper presents a deep learning network that performs automatic detection of defects by inspecting full ultrasonic guided wave signals excited in plate structures. The findings show that the algorithm, which is an adaptation of WaveNet, and hence is based on causal dilated convolutional neural networks, is effectively able to learn features and/or patterns related to the presence of waves scattered from damage, thus eliminating the need for any feature engineering to be performed by human operators. The network outperformed the widely used conventional approach that combines the optimal baseline selection and baseline signal stretch compensation methods when tested on two different datasets. The first dataset consisted of finite element simulated Lamb wave signals acquired in a pitch-catch configuration on a steel plate across a 50 °C range of temperature variations, and the second was a publicly available experimental dataset of Lamb wave signals also acquired in pitch-catch mode on a composite plate with a 40 °C range of variations. The improvements over the conventional approach are particularly encouraging when analyzing signals at temperatures well outside the temperature range available in the set of baseline signals, hence suggesting that this class of algorithms can complement or substitute existing methods, especially when testing occurs at unseen environmental and operational conditions, or when the effects of sensor drift make conventional methods less effective. [ABSTRACT FROM AUTHOR]

Quality of experience (QoE) assessment for adaptive video streaming plays a significant role in advanced network management systems. It is especially challenging in case of dynamic adaptive streaming schemes over HTTP (DASH) which has increasingly complex characteristics including additional playback issues. In this paper, we provide a brief overview of adaptive video streaming quality assessment. Upon our review of related works, we analyze and compare different variations of objective QoE assessment models with or without using machine learning techniques for adaptive video streaming. Through the performance analysis, we observe that hybrid models perform better than both quality-of-service (QoS) driven QoE approaches and signal fidelity measurement. Moreover, the machine learning-based model slightly outperforms the model without using machine learning for the same setting. In addition, we find that existing video streaming QoE assessment models still have limited performance, which makes it difficult to be applied in practical communication systems. Therefore, based on the success of deep learned feature representations for traditional video quality prediction, we also apply the off-the-shelf deep convolutional neural network (DCNN) to evaluate the perceptual quality of streaming videos, where the spatio-temporal properties of streaming videos are taken into consideration. Experiments demonstrate its superiority, which sheds light on the future development of specifically designed deep learning frameworks for adaptive video streaming quality assessment. We believe this survey can serve as a guideline for QoE assessment of adaptive video streaming. • Quality of experience assessment for adaptive video streaming plays a key role in multimedia communication systems. • Several subjective quality assessment databases proposed for adaptive video streaming are reviewed. • Comparisons of existing state-of-the-art objective quality assessment models for streaming videos are conducted. • Applying the deep feature representations based on spatio-temporal human visual perception can deliver promising results. • The experiments shed light on the future development of specifically designed deep learning frameworks for adaptive video streaming quality assessment. [ABSTRACT FROM AUTHOR]

This paper develops a new nonlinear transformation-based augmentation method for Convolutional Neural Network (CNN) approach with vibration signals of simple, small scale and elementary reference models for the classification or prediction of vibration signals of perplex healthy or damaged systems using a smart diagnosis system. The accuracy of deep learning algorithm being highly dependent on the quantity of qualified data, the acquiring of a large set of formatted data for the training and verification of a deep learning algorithm is essential. Unfortunately, many scientific and engineering application domains do not allow access to a Big Data accurately bearing domain knowledge and the artificial intelligent (AI) based classification methods suffer from the lack of data and often end up with poor prediction. To overcome this issue, data augmentation approaches have been utilized. In many applications, however, the obtaining of data reflecting physical phenomena is even not possible. To overcome this issue, this research suggests a new nonlinear transformation-based augmentation approach mapping from the data obtained from lab-scale healthy models to the data of complex real healthy models whose data in the damaged status are hard to be obtained. The nonlinear transformation method defined between the data of lab-scale healthy models and the data of complex real healthy model is then applied to predict the data of complex real damaged models for an accurate classification. To validate the concept of the nonlinear transformation augmentation, several vibration examples including an example showing the mode switching are considered. To extract discriminating features from the vibration-based spectrograms using a deep learning algorithm, the nonlinear transformation-based augmentation and the classification between healthy and damaged structures are presented. • A new nonlinear mapping augmentation approach is presented for vibration signals. • The vibration signals of simple models can be augmented with the augmentation approach. • Damage signals of complex real models are predicted with the augmentation approach. • The augmentation approach can consider the mode switching phenomenon. • Various fracture features can be predicted with the augmentation approach. [ABSTRACT FROM AUTHOR]

A few decades ago, the classification of ships by means of their acoustics signature depended on the trained ears of SONAR system operators. With the advances of machine and deep learning techniques, new methods of classification were proposed in order to assist the operator to identify and classify the acoustic event. In this scenario, this paper presents a passive SONAR classification method based on Cyclostationary Analysis and Convolutional Neural Networks. The Spectral Coherence of the ship-radiated noise was extract, and served as input for a modified MobileNetV2 Convolutional Neural Network classifier. The method was benchmarked in two different dataset and was able to discriminate different classes. In parallel, another three classifiers based on Support Vector Machine were developed. Comparison with previous articles that used the same database was made, as well as, an evaluation of noise immunity was conducted. It was observed that the proposed method presented higher accuracy, even for low signal-to-noise ratios. • A ship classification method based on CNN and Cyclostationarity Analysis is proposed. • It uses the Spectral Coherence as a feature extraction for ship-radiated noise. • The MobileNetV2 CNN is used to classify vessels through the Spectral Coherence graph. • An evaluation of noise immunity of the classifier's accuracy is conduct. • The developed methodology is tested on two datasets obtaining good performance. [ABSTRACT FROM AUTHOR]

• A new DNN model, DMCNet is proposed for feature learning from vibration signal. • A morphological layer is utilized to extract impulses and filter out noise on signals. • A kurtosis-based weighted fusion method is proposed for morphological feature selection. • Recalibrated residual learning is developed to perform feature learning from morphological features. Vibration signals are utilized widely for machinery fault diagnosis. These typical deep neural networks (DNNs), e.g., convolutional neural networks (CNNs) perform well in feature learning of vibration signals for machinery fault diagnosis. However, vibration signals collected from real industry are often noised and nonstationary. It is still a challenging problem for CNNs to implement feature extraction and noise reduction from a large number of vibration signals. In this paper, a novel DNN, deep morphological convolutional network (DMCNet) is proposed for feature learning of gearbox vibration signals. Firstly, two parallel branches with different morphological operations (i.e., opening and closing) are adopted to filter out the noise and perform feature mapping on vibration signals. The structure elements of morphological operation can be updated in the network training by using the back-propagation learning algorithm. Secondly, a kurtosis-based feature fusion method is proposed to enhance channels with strong impulsive features. Finally, a recalibrated residual learning is further proposed for feature learning on the morphological feature maps, where a recalibrated skip connection is used for feature selection. The experimental results indicate that DMCNet can implement feature leaning of gearbox vibration signals well and outperform those state-of-the-art CNNs. [ABSTRACT FROM AUTHOR]

• Recognition of local cortical folding patterns is hard due to their high variability. • Here, three new methods are proposed for their automatic classification. • They are tested on the Anterior Cingulate Cortex patterns and the Power Button Sign. • The CNN-based model is faster to use and has similar performance to the other models. • The two other models seem more effective in managing unbalanced problems. The study of local cortical folding patterns showed links with psychiatric illnesses as well as cognitive functions. Despite the tools now available to visualize cortical folds in 3D, manually classifying local sulcal patterns is a time-consuming and tedious task. In fact, 3D visualization of folds helps experts to identify different sulcal patterns but fold variability is so high that the distinction between these patterns sometimes requires the definition of complex criteria, making manual classification difficult and not reliable. However, the assessment of the impact of these patterns on the functional organization of the cortex could benefit from the study of large databases, especially when studying rare patterns. In this paper, several algorithms for the automatic classification of fold patterns are proposed to allow morphological studies to be extended and confirmed on such large databases. Three methods are proposed, the first based on a Support Vector Machine (SVM) classifier, the second on the Scoring by Non-local Image Patch Estimator (SNIPE) approach and the third based on a 3D Convolution Neural Network (CNN). These methods are generic enough to be applicable to a wide range of folding patterns. They are tested on two types of patterns for which there is currently no method to automatically identify them: the Anterior Cingulate Cortex (ACC) patterns and the Power Button Sign (PBS). The two ACC patterns are almost equally present whereas PBS is a particularly rare pattern in the general population. The three models proposed achieve balanced accuracies of approximately 80% for ACC patterns classification and 60% for PBS classification. The CNN-based model is more interesting for the classification of ACC patterns thanks to its rapid execution. However, SVM and SNIPE-based models are more effective in managing unbalanced problems such as PBS recognition. [ABSTRACT FROM AUTHOR]