Search

As an all-in-one process that blended design, materials, and manufacturing, additive manufacturing (AM) adopted a bottom-up stacking approach, thereby avoiding excessive infrastructure and complex production steps. However, the different process stages of AM were influenced by multi-scale factors such as material characteristics, parameter settings, and process stability, resulting in uncertainty. Therefore, the effective containment and accurate evaluation of automated production processes were crucial. With the rapid development and widespread application of artificial intelligence (AI) technologies in the Industry 4.0 era, computer vision (CV) techniques, as an important component, could achieve multi-level understanding of the real world through traditional machine learning (ML) and efficient deep learning (DL) approaches based on information-dense cross-media data such as images, videos, and three-dimensional models. Hence, the organic integration of CV and AM was of great significance for the virtual simulation, global perception, real-time monitoring, and quality measurement of the machining processes. According to the chronological order and logical relationship of manufacturing engineering, this paper organized and summarized the application of CV techniques in material characteristics perception and stacking process simulation before the implementation of projects, in situ inspection, and law exploration during the manufacturing processes, as well as product evaluation and defect detection after processing, which demonstrated its powerful performance and great potential in the intelligence and integration of AM. [ABSTRACT FROM AUTHOR]

Suicide is a significant public health issue that devastates individuals and society. Early warning systems are crucial in preventing suicide. The purpose of this research is to create a deep learning model to identify suicide risk using a combination of bidirectional encoder representations from transformers (BERT) and an ensemble of multiple convolutional neural networks (CNN). BERT is used to encode the text data into numerical representations that capture the context-aware meaning of words and phrases, while ensemble CNN is used to analyze the encoded text data to identify patterns and relationships relevant to suicide risk. The model is trained on a large corpus of text data from social media and suicide notes and evaluated on a validation set of labeled data. The results of the proposed model perform better than the recent approaches in detecting suicide risk and can potentially be a valuable tool for suicide prevention efforts. [ABSTRACT FROM AUTHOR]

Deep learning has proven to be highly effective for human activity recognition (HAR) when large amount of labelled data is available for the target task. However, training a deep learning model to generalize well on a new task with just-few observations of labelled data is an active area of research. In this paper, a novel few-shot transfer learning (FSTL) approach is proposed for classification of human activities using just few instances (shots) of the data obtained from a wearable system assembled to collect inertial sensor data for different human activities, performed by two users. First, a deep learning model is trained on a large publicly available HAR dataset. The model parameters of such a model are then fine-tuned using the Reptile algorithm to determine the optimal initial parameter set using which, the model will classify activities with just few-shots of data from the target task. The proposed FSTL approach yields an average classification accuracy of 74.86 ± 0.71% and 79.20 ± 1.05% for 3-way, 5-shot classification of new activities performed by a single user and same set of activities performed by a new user, respectively. When the pre-trained weights are used as the initial weights in the Reptile algorithm, the generalization ability of the model improves by about 10% for 3-way, 5-shot classification as compared to using few-shot learning without parameter transfer. [ABSTRACT FROM AUTHOR]

Vision-based Human Action Recognition is difficult owing to the variations in the same action performed by various people, the temporal variations in actions, and the difference in viewing angles. Researchers have recently adopted multi-modal visual data fusion strategies to address the limitations of single-modality methodologies. Many researchers strive to produce more discriminative features because most existing techniques' success relies on feature representation in the data modality under consideration. Human action consists of several sub-actions whose duration vary between individuals. This paper proposes a multifarious learning framework employing action data in depth and skeleton formats. Firstly, a novel action representation named Multiple Sub-action Enhanced Depth Motion Map (MS-EDMM), integrating depth features from overlapping sub-actions, is proposed. Secondly, an efficient method is introduced for extracting spatio-temporal features from skeleton data. This is achieved by dividing the skeleton sequence into sub-actions and summarizing skeleton joint information for five distinct human body regions. Next, a multi-stream deep learning model with Attention-guided CNN and residual LSTM is proposed for classification, followed by several score fusion operations to reap the benefits of streams trained with multiple data types. The proposed method demonstrated a superior performance of 1.62% over an existing method that utilized skeleton and depth data, achieving an accuracy 89.76% on a single-view UTD-MHAD dataset. Furthermore, on the multi-view NTU RGB+D dataset demonstrated encouraging performance with an accuracy of 89.75% in cross-view and 83.8% in cross-subject evaluations. [ABSTRACT FROM AUTHOR]

Rapid detection of damages occurring as a result of natural disasters is vital for emergency response. In recent years, remote sensing techniques have been commonly used for the automatic categorization and localization of such events using satellite images. Trained based on natural disaster images, a convolutional neural network (CNN) has been applied as a highly successful method, with its ability to reveal outstanding features. Studies aiming to detect target points obtained as a result of extracting visual features from natural images within these networks have achieved their goals. In this study, ensemble learning methods have been suggested as a means to develop the detection of landslide areas from landslide satellite images. Landslide image dataset has been trained for their categorization in CNN models and then they have been used again to localize landslide regions. While model predictions develop overall performance and status, different ensemble strategies have been used and integrated to reduce the sensitivity to prediction variance and training data. Class-selective relevance mapping (CRM) has been used to visualize individual CNN models and ensemble learned behaviors. As a result of the comparisons made based on mean average precision metrics and the criteria of intersection over union, model ensembles have proved to show higher localization performance than any other individual model. [ABSTRACT FROM AUTHOR]

Tensile testing (aka tension testing) is a widely employed mechanical testing technique for analyzing materials' properties and behavior under applied stress. Tensile testing plays a pivotal role in helping engineers to make informed decision about material selection and usage. Despite its importance, there is a limited numbers of studies that explored the potential of AI techniques for real time monitoring and material behavior prediction in tensile testing. To this end, this work presents a deep learning model designed to predict the material's condition throughout tensile testing and provide an early warning prior to fracture. By leveraging a comprehensive dataset of tension test video samples, the proposed model utilizes both convolution and recurrent neural networks to extract pertinent spatial and temporal visual features, thereby predicting the frames at which material deformation and fracture occur. The evaluation results of our research showed that the proposed model achieved a predictive ability with an F1-score of 97%, on average. The implications of our research are significant for industries and researchers in the field of materials science and engineering. By accurately predicting material status, our model enables autonomous, real time analysis of material behavior during tensile testing, leading to better time and cost efficiency in various applications. [ABSTRACT FROM AUTHOR]

Visual XAI methods enable experts to reveal importance maps highlighting intended classes over input images. This research paper presents a novel approach to visual explainable artificial intelligence (XAI) for object detection in deep learning models. The study investigates the effectiveness of activation maps generated by five different methods, namely GradCAM, GradCAM++, EigenCAM, HayCAM, and a newly proposed method called "HayCAMJ", in detecting objects within images. The experiments were conducted on two datasets (Pascal VOC 2007 and Pascal VOC 2012) and three models (ResNet18, ResNet34, and MobileNet). Zero padding was applied to resize and center the objects due to the large objects in the images. The results show that HayCAMJ performs better than other XAI techniques in detecting small objects. This finding suggests that HayCAMJ has the potential to become a promising new approach for object detection in deep classification models. [ABSTRACT FROM AUTHOR]

Brain–Computer Interface is tested as a successful method in improving human cognitive functions such as attention and memory. Attention plays a significant role in areas ranging from a person's day-to-day life to educational domain and professional activities. When attention is evaluated using camera-based techniques, the users may suffer privacy issues. Using Brain–Computer Interface (BCI) to obtain a measure of attention will be useful in this regard. The paper proposes a Human Attention Recognition System (HARS) in which EEG signal acquisition is used to obtain the attention of the individual, Renyi's entropy-based mutual information method is used for feature selection and a deep learning-based classifier is used to classify the signals. HADS is not camera-based: therefore, faces of the subjects are not revealed. EEG signals were collected using the Ultracortex Mark III dry electrodes and were visualised using OpenBCI GUI (Graphical User Interface). The experiment is validated using the publicly available Confused Student EEG dataset from Kaggle, giving an accuracy of 99.21%. The results indicate that the proposed method can identify attention levels accurately and can be effectively used in educational systems, biofeedback and medical research. [ABSTRACT FROM AUTHOR]

Point clouds in the real world are often sparse and incomplete. Point cloud completion aims to restore incomplete point clouds into meaningful shapes. In recent years, point cloud completion has attracted the interest of many researchers. In most existing methods, the efficiency of completion is generally ignored in favor of producing meaningful shapes with adequate details. This paper proposes a fast point completion network (FPCN). FPCN is mainly composed of a multi-scale attention encoder (MSAE) and a structural refinement (SR) module. MSAE first obtains multi-scale geometric information by extracting incomplete inputs at different resolutions. After that, MSAE fuses multi-scale geometric information through cross-attention mechanisms. Compared with existing mainstream encoders, MSAE can extract rich geometric information from input point clouds with low complexity. The SR module aims to extract local information from the coarse point clouds to guide the process of extending points. Furthermore, the process of extending points is achieved by a replication strategy. Compared to existing folding-based decoders, the SR module can produce fine point clouds with more local details. Compared to existing transformer-based decoders, the SR module has a lower calculation price by employing a replication strategy to generate high-resolution point clouds. In conclusion, FPCN can restore partial point clouds into meaningful shapes efficiently, and the outcomes of completion contain sufficient details. Extensive experiments on various datasets demonstrate the performance and efficiency of the FPCN in point cloud completion. Source code is available at https://github.com/doldolOuO/FPCN. [ABSTRACT FROM AUTHOR]

This study introduces an online (supervised) learning method to design nonlinear auto-regressive moving average (NARMA) controllers for feedback-linearized nonlinear single-input single-output (SISO) systems. The algorithm ensures Schur stability of the overall closed-loop system and provides adaptiveness and robustness for the NARMA controllers. The first stage of the method derives, in a data-dependent way, a feedback-linearized model of the nonlinear plant by using its input and output sample pairs. The method's second stage, which constitutes the novel part of the presented study, builds up an online learning scheme for the linear auto-regressive moving average (ARMA) controller based on an already learned feedback-linearized model of the nonlinear plant. During online supervised learning, ARMA parameters of the feedback-linearized SISO plant model and the closed-loop ARMA model are computed by minimizing the plant identification and the closed-loop system tracking errors. Both errors are defined as ℓ 1 , ε , namely ε-insensitive loss functions that provide NARMA controller the robustness against noise and outliers. The proposed online learning control algorithm is applied to a rotary inverted pendulum model and to a real rotary inverted pendulum setup. The tracking performance of the developed controller is compared with those of the linear quadratic regulator and coupled sliding mode controller in terms of mean square error. [ABSTRACT FROM AUTHOR]

The increasing reliance on internet-based services has heightened the vulnerability of network infrastructure to cyberattacks, particularly distributed denial of service (DDoS) attacks. These attacks can cause severe disruptions and significant financial losses. Early detection of malicious traffic is crucial in effectively combating such threats. This paper presents an innovative approach called the Encoder-Stacked deep neural networks (E-SDNN) model, which leverages Stacked/bagged multi-layer perceptrons (MLP) for accurate DDoS attack detection. The proposed method employs an encoder to select pertinent features from a preprocessed dataset, enabling precise attack detection. Extensive experiments were conducted on benchmark cybersecurity datasets, namely CICDS2017 and CICDDoS2019, encompassing various DDoS attack scenarios. The experimental results demonstrate the superiority of the E-SDNN model compared to state-of-the-art methods. The proposed E-SDNN model achieved an impressive overall accuracy rate of 99.94% and 98.86% for CICDDS2017 and CICDDoS2019, respectively. [ABSTRACT FROM AUTHOR]

Limitations in computer-assisted diagnosis include lack of labeled data and inability to model the relation between what experts see and what computers learn. Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their accountability and transparency level must be improved to transfer this success into clinical practice. The reliability of machine learning decisions must be explained and interpreted, especially for supporting the medical diagnosis. While deep learning techniques are broad so that unseen information might help learn patterns of interest, human insights to describe objects of interest help in decision-making. This paper proposes a novel approach, DeepCraftFuse, to address the challenge of combining information provided by deep networks with visual-based features to significantly enhance the correct identification of cancerous tissues in patients affected with Barrett's esophagus (BE). We demonstrate that DeepCraftFuse outperforms state-of-the-art techniques on private and public datasets, reaching results of around 95% when distinguishing patients affected by BE that is either positive or negative to esophageal cancer. [ABSTRACT FROM AUTHOR]

Natural disasters can be unpredictable and catastrophic. Even after the event, the repercussions are prolonged due to the incompetence of disaster management strategies. To mitigate the effects of a natural hazard, disaster management teams have to rapidly come forth with innovative plans of action. To contain the damage that a disaster causes, the response time, and preparedness of a disaster management team is crucial. Having a forewarning about the nature of the disaster can prove to be beneficial for the management team. Remote/unreachable areas such as deep parts of the forests, far-flung rural areas, oceans and other hard to reach locations are at a higher risk of receiving poorer aid and response due to the lack of communication and connectivity with the rest of the world. Our paper provides a solution for this particular problem by suggesting UAV/Drones for surveillance and monitoring in these inaccessible, disaster struck areas. The drones not only monitor the situation but constantly click pictures and send them to a base station, where the images are used to collect insights about the type of disaster that has to be dealt with. The paper proposes a deep learning model based on feature concatenation for classification of disasters which can be deployed at the base station, and can receive data in the form of images from a UAV/drone hovering over the affected place. The proposed model is efficient and able to achieve a higher accuracy as compared to the leading CNN models and closely related recent works as well. [ABSTRACT FROM AUTHOR]

This paper proposes a new method to improve the training efficiency of deep convolutional neural networks. During training, the method evaluates scores to measure how much each layer's parameters change and whether the layer will continue learning or not. Based on these scores, the network is scaled down such that the number of parameters to be learned is reduced, yielding a speed-up in training. Unlike state-of-the-art methods that try to compress the network to be used in the inference phase or to limit the number of operations performed in the back-propagation phase, the proposed method is novel in that it focuses on reducing the number of operations performed by the network in the forward propagation during training. The proposed training strategy has been validated on two widely used architecture families: VGG and ResNet. Experiments on MNIST, CIFAR-10 and Imagenette show that, with the proposed method, the training time of the models is more than halved without significantly impacting accuracy. The FLOPs reduction in the forward propagation during training ranges from 17.83% for VGG-11 to 83.74% for ResNet-152. As for the accuracy, the impact depends on the depth of the model and the decrease is between 0.26% and 2.38% for VGGs and between 0.4 and 3.2% for ResNets. These results demonstrate the effectiveness of the proposed technique in speeding up learning of CNNs. The technique will be especially useful in applications where fine-tuning or online training of convolutional models is required, for instance because data arrive sequentially. [ABSTRACT FROM AUTHOR]

Ensuring the safety of workers and machinery during operations is a critical task in the construction sites. However, an inevitable circumstance in construction sites is the complex and dynamic environment, which often leads to occlusions. When detecting occluded objects in construction sites, general vision-based approaches tend to exhibit lower accuracy and may even miss detections, resulting in potential safety hazards. To handle this issue, this paper proposes a vision-based approach for detecting occluded objects in construction sites. Firstly, the proposed detection algorithm adopts the state-of-the-art YOLOv7 as its backbone. To enhance its capability in capturing contextual information of occluded objects, a novel channel attention mechanism is employed. Then, a design scheme for the detector head is provided by integrating a novel loss function Scylla-Intersection over Union (SIoU) and the non-maximum suppression (NMS) strategy. With the help of the loss function SIoU, the network can compute the loss values of occluded objects more accurately. To ensure that the network can select the right predicted box which closely aligns with the ground truth, the Euclidean distance is utilized as spatial penalty factor during the NMS stage. By implementing these two strategies, the proposed method can preserve both the category information and bounding boxes of occluded objects, which makes them possible to be detected. Finally, detailed experiments are done to verify the proposed method. Experimental results demonstrate that the proposed method has the potential for improving the detection accuracy. Moreover, it shows a better performance in detecting occluded objects in the dynamic construction sites compared to the existing baselines. [ABSTRACT FROM AUTHOR]

Aluminum foam is a multifunctional material with excellent shock absorption and heat-insulating properties and is expected to be used in many industrial fields. Since aluminum foam is fabricated by foaming aluminum, variations in pore structures can be observed even when foaming is conducted under the same conditions. Therefore, the establishment of a quality assurance method for the properties of aluminum foam as a product is a major issue. In this study, we attempted to estimate the plateau stress of aluminum foam by machine learning using only X-ray computed tomography (CT) images. A supervised learning neural network model was created using the obtained X-ray CT images of the entire three-dimensional structure and a set of plateau stresses obtained from actual compression tests of the corresponding aluminum foam. Using the created model, we showed that the plateau stress of a new aluminum foam compression specimen, which was different from the aluminum foam compression specimens used for training, can be estimated from X-ray CT images only. That is, it was shown that the mechanical properties of materials with complex geometries, such as foam, can be predicted nondestructively. In the future, it is expected to predict the mechanical properties of the foam by X-ray transmission and visual photographs, which are easier to obtain. We also showed that as the resolution of X-ray CT images improved, the mean error became smaller, namely, the estimation accuracy improved. [ABSTRACT FROM AUTHOR]

As the integration of mechanical engineering and deep learning fields becomes increasingly intertwined, the application of experimental thermal error modeling in intelligent manufacturing has gained significant importance. In this paper, the issue of spindle thermal error is treated as a multivariate time series problem due to the thermal transfer characteristics. This study aims to address the challenge of modeling intermittent multivariate time series spindle thermal errors under a wide range of environmental temperatures and various operational scenarios. To tackle this challenge, a substantial volume of experimental data, capable of effectively reflecting the patterns of spindle thermal error variations, was collected through experiments conducted at multiple speeds and under various operational scenarios. Subsequently, the acquired thermal error data underwent intermittent multivariate time series transformation (IMTS) to suit the serialized deep learning model. The study introduces the Crossformer model into the field of thermal error modeling for the first time, which is a variant of the Transformer model. The Crossformer model exhibits remarkable adaptability to temporal aspects while effectively maintaining its focus on data features. Ultimately, this study resulted in the development of the IMTS-CrossformerR experimental thermal error model. Throughout the research, a comprehensive examination of various models was undertaken, including two traditional Transformer models and other thermal error deep learning and machine learning models. The results indicate that the proposed model outperforms its counterparts across multiple model metrics and predictive capabilities. Particularly noteworthy is its substantial improvement in the range (± 5) ratio of residual fluctuations reaching 95.7%, a key engineering metric. [ABSTRACT FROM AUTHOR]

One of the main challenges that prevent wide adoption of Single Point Incremental Forming (SPIF) is the geometric accuracy of the process resulting primarily from the effect of springback. There are various expedients that can be adopted to address this, but one of the most common is tool path correction. The challenge is then how best to predict springback so as to implement tool path correction. It is established that springback, to a large extent, is related to the geometry of the part to be manufactured. The proposed mechanism uses a novel point series representation to capture local geometries that then form a global bank of geometries for general use. Each point series can then be associated with a predicted springback value generated using deep or machine learning. Experiments are reported using a Long Short Term Memory (LSTM) model coupled with a Multilayer Perception Network (MLP), and a Support Vector Machine (SVM) regression model. A best R2, "Coefficient of Determination", of 0.9181 was obtained indicating that the proposed approach provided a realistic solution to the current limitations of SPIF. [ABSTRACT FROM AUTHOR]

Over the preceding decade, there has been a discernible surge in the prominence of artificial intelligence, marked by the development of various methodologies, among which deep learning emerges as a particularly auspicious technique. The captivating attribute of deep learning, characterised by its capacity to glean intricate feature representations from data, has served as a catalyst for pioneering approaches and methodologies spanning a multitude of domains. In the face of the burgeoning exponential growth in digital medical image data, the exigency for adept image compression methodologies has become increasingly pronounced. These methodologies are designed to preserve bandwidth and storage resources, thereby ensuring the seamless and efficient transmission of data within medical applications. The critical nature of medical image compression accentuates the imperative to confront the challenges precipitated by the escalating deluge of medical image data. This review paper undertakes a comprehensive examination of medical image compression, with a predominant focus on sophisticated, research-driven deep learning techniques. It delves into a spectrum of approaches, encompassing the amalgamation of deep learning with conventional compression algorithms and the application of deep learning to enhance compression quality. Additionally, the review endeavours to explicate these fundamental concepts, elucidating their inherent characteristics, merits, and limitations. [ABSTRACT FROM AUTHOR]

Multi-Object Tracking, also known as Multi-Target Tracking, is an important area of computer vision with various applications in different domains. The advent of deep learning has had a profound impact on this field, forcing researchers to explore innovative avenues. Deep learning methods have become the cornerstone of today's state-of-the-art solutions, consistently delivering exceptional tracking results. However, the significant computational demands of deep learning models require powerful hardware resources that do not always match real-time tracking requirements, limiting their practical applicability in real-world scenarios. Thus, there is an imperative to strike a balance by merging robust deep learning strategies with conventional approaches to enable more accessible, cost-effective solutions that meet real-time requirements. This paper embarks on this endeavor by presenting a hybrid strategy for real-time multi-target tracking. It effectively combines a classical optical flow algorithm with a deep learning architecture tailored for human crowd tracking systems. This hybrid approach achieves a commendable balance between tracking accuracy and computational efficiency. The proposed architecture, subjected to extensive experimentation in various settings, demonstrated notable results, achieving a Mean Object Tracking Accuracy (MOTA) of 0.608. This level of performance placed it as the highest ranking solution on the MOT15 benchmark, surpassing the state-of-the-art benchmark of 0.549, and consistently ranked among the superior models on the MOT17 and MOT20 benchmarks. Additionally, the incorporation of the optical flow phase resulted in a substantial reduction in processing time, nearly halving the duration, while simultaneously maintaining accuracy levels comparable to established techniques. [ABSTRACT FROM AUTHOR]

The data-driven modular deep fuzzy model has demonstrated excellent forecasting performance due to its clear architecture and powerful fuzzy inference ability. However, the fixed structure predesigned for specific types of datasets limits the further improvement of generalization ability, especially when facing high-dimensional datasets with the challenging problem of rule explosion, thereby suffering from the time-consuming training process. To overcome these limitations, a novel randomized fuzzy modules (RFMs) stacked residual deep fuzzy system (RFM-RDFS) is proposed in this study. Firstly, to obtain high-precision outputs, this paper presents a residual deep fuzzy framework by stacking RFMs layer by layer, where the modules in the current layer only learn the residual of the previous layer. Secondly, to improve the design freedom for different applications, the RFM with multiple inputs is raised, where the input variables are allocated by a sliding window along the input spaces of the layers, so the number of the parallel modules in the same layer can be flexibly adjusted. Meanwhile, in the RFM, inspired by the extreme learning machine, the centers and widths of Gaussian membership functions are randomly generated, which can greatly accelerate the training speed. Thirdly, to strengthen the approximation performance, the consequent parameters of the fuzzy rules in each RFM are learned by minimizing the loss function with the regularization terms through the ridge regression algorithm. Finally, extensive experiments on various datasets are conducted to compare the proposed RFM-RDFS with several state-of-the-art shallow and deep models. The detailed comparison results clearly show that the proposed RFM-RDFS is more flexible and compact, and has greater advantages in terms of the interpretability, generalization ability and robustness. [ABSTRACT FROM AUTHOR]

Graph neural networks (GNNs) are reshaping our understanding of biomedicine and diseases by revealing the deep connections among genes and cells. As both algorithmic and biomedical technologies have advanced significantly, we are entering a transformative phase of personalized medicine. While pioneering tools like graph attention networks (GATs) and graph convolutional neural networks (Graph CNN) are advancing graph-based learning, the rise of single-cell sequencing techniques is reshaping our insights on cellular diversity and function. Numerous studies have combined GNNs with single-cell data, showing promising results. In this work, we highlight the GNN methodologies tailored for single-cell data over the recent years. We outline the diverse range of graph deep learning architectures that center on GAT methodologies. Furthermore, we underscore the several objectives of GNN strategies in single-cell data contexts, ranging from cell-type annotation, data integration and imputation, gene regulatory network reconstruction, clustering and many others. This review anticipates a future where GNNs become central to single-cell analysis efforts, particularly as vast omics datasets are continuously generated and the interconnectedness of cells and genes enhances our depth of knowledge in biomedicine. [ABSTRACT FROM AUTHOR]

With the rapid development of data acquisition and storage technology, spatio-temporal (ST) data in various fields are growing explosively, so many ST prediction methods have emerged. The review presented in this paper mainly studies the prediction of ST series. We propose a new taxonomy organized along three dimensions: ST series prediction methods (focusing on time feature learning, focusing on spatial feature learning, and focusing on spatial–temporal feature learning), techniques of ST series prediction (the RNN-, CNN-, and transformer-based models, as well as the CNN-based-composite model and GNN-based-composite models, and the miscellaneous model) and ST series prediction results (single target and multi-target). We first introduce and explain each dimension of the taxonomy in detail. After providing this three-dimensional view, we comprehensively review and compare the recent related ideas in the literature and analyze their advantages and limitations. Moreover, we summarize the key information of the existing literature and provide guidance for researchers to select suitable models. Second, we summarize the different applications of deep learning models in ST series prediction based on current literature and list relevant datasets and download links per application classifications. Lastly, we comprehensively analyze the current innovation and challenges and suggest future directions for researching ST series prediction after comparing and analyzing the computing performance of these forecasting models. In addition, each method or model solves one aspect of the challenge, which means that two or more methods should be combined to solve more challenges at the same time. We hope this article provides readers a broader and deeper understanding of the field of ST series research. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic has profoundly impacted healthcare systems and economies worldwide, leading to the implementation of travel restrictions and social measures. Efforts such as vaccination campaigns, testing, and surveillance have played a crucial role in containing the spread of the virus and safeguarding public health. There needs to be more research exploring the transmission dynamics of COVID-19, particularly within European nations. Therefore, the primary objective of this research was to examine the spread patterns of COVID-19 across various European countries. Doing so makes it possible to implement preventive measures, allocate resources, and optimize treatment strategies based on projected case and mortality rates. For this purpose, a hybrid prediction model combining CNN and LSTM models was developed. The performance of this hybrid model was compared against several other models, including CNN, k-NN, LR, LSTM, MLP, RF, SVM, and XGBoost. The empirical findings revealed that the CNN-LSTM hybrid model exhibited superior performance compared to alternative models in effectively predicting the transmission of COVID-19 within European nations. Furthermore, examining the peak of case and death dates provided insights into the dynamics of COVID-19 transmission among European countries. Chord diagrams were drawn to analyze the inter-country transmission patterns of COVID-19 over 5-day and 14-day intervals. [ABSTRACT FROM AUTHOR]

Parkinson's disease (PD) is a condition that degenerates over time and impairs speech and pronunciation because brain cells have died. This research work aims to predict parkinson disease using the voice features extracted from speech signals recorded from PD individuals with dysphonic speech disorders by employing deep learning algorithms. PD is challenging to diagnose early on in the clinical presentation. To address the issue in machine learning methods, this paper proposes a neural network model by processing speech signals to classify PD using the University of California Irvine (UCI) machine learning repository dataset. Initially, a pre-loss reduction module is created by using pre-sampling to make the dataset balanced by reducing the dimensionality and maintaining the size of the space without influencing the learning process for data preparation. The relevant features are derived using a novel multi-agent salp swarm (MASS) algorithm, and a novel Parkinson classification neural network (PCNN) is proposed to classify Parkinson's patients with high accuracy employing these derived features. The result shows that the models that use MASS-PCNN produce higher classification accuracy of 99.1%, precision of 97.8%, recall of 94.7% and F1-score of 0.995 when paralleled to the existing models. As an outcome, the suggested model will perform superior to common convolutional neural networks. [ABSTRACT FROM AUTHOR]

On most e-commerce websites, there are two crucial factors that customers rely on to assess product quality and dependability: customer reviews provided online and related ratings. Reviews offer feedback to customers about the product's merits, reasons for negative reviews, and feelings of satisfaction or dissatisfaction with the provided service. As for ratings, they express customer opinions about the product's quality as numerical values from one to five (one or two for the worst opinion, three for the neutral opinion, and four or five for the best opinion). Usually, the customer reviews may be inconsistent with their relevant ratings; the customer may write the worst review despite providing a four- or five-star rating or write the best review with only a one- or two-star rating. Due to this inconsistency, customers may need help to identify relevant information. Therefore, it is required to develop a model that can classify reviews as either positive or negative, depending on the polarity of thoughts, to demonstrate if there is an inconsistency between customer reviews and their actual ratings by comparing them with the ratings resulting from the model. This paper proposes an efficient deep learning (DL) model for classifying customer reviews and assessing whether there is inconsistency. The recommended model's performance and stability are examined on a large dataset of product reviews from Amazon e-commerce. The experimental findings showed that the proposed model dominates and significantly outperforms its peers regarding prediction accuracy and other performance measures. [ABSTRACT FROM AUTHOR]

The growing prevalence of deep neural networks (DNNs) across various fields raises concerns about their increasing energy consumption, especially in large data center applications. Identifying the best combination of optimization techniques to achieve maximum energy efficiency while maintaining system performance is challenging due to the vast number of techniques available, their complex interplay, and the rigorous evaluation required to assess their impact on the model. To address this gap, we propose an open-source methodological framework for the systematic study of the influence of various optimization techniques on diverse tasks and datasets. The goal is to automate experimentation, addressing common pitfalls and inefficiencies of trial and error, saving time, and allowing fair and reliable comparisons. The methodology includes model training, automatic application of optimizations, export of the model to a production-ready format, and pre- and post-optimization energy consumption and performance evaluation at inference time using various batch sizes. As a novelty, the framework provides pre-configured "optimization strategies" for combining state-of-the-art optimization techniques that can be systematically evaluated to determine the most effective strategy based on real-time energy consumption and performance feedback throughout the model life cycle. As an additional novelty, "optimization profiles" allow the selection of the optimal strategy for a specific application, considering user preferences regarding the trade-off between energy efficiency and performance. Validated through an empirical study on a DNN-based cyber threat detector, the framework demonstrates up to 82% reduction in energy consumption during inference with minimal accuracy loss. [ABSTRACT FROM AUTHOR]

The prognostic analysis for high grade serous adenocarcinoma (HGSC) holds significant clinical importance. However, current prognostic analysis primarily relies on statistical techniques like logistic regression and chi-square analysis alongside traditional machine learning methods based on pattern recognition. These approaches face challenges in addressing the limited reliability and validity of evaluation results, as well as the absence of reliable prognostic indicators. To identify a reliable prognostic evaluation method for high grade serous adenocarcinoma, a novel prognostic evaluation method was constructed using multi-modal deep learning techniques and compared with existing methods using data from 210 patients with high grade serous adenocarcinoma (stage III). The experimental results showed that the accuracy of this method for prognostic analysis was 80.0%, and the detection rate for poor prognosis cases was 82.87%, which was superior to current methods. Our proposed method could also automatically extract key features from different datasets and efficiently predict patient outcomes. Overall, this study laid the groundwork to overcome the difficulties in the prognostic evaluation of HGSC, help clinicians better understand the pathogenesis, and improve the long-term survival rates of this patient population. [ABSTRACT FROM AUTHOR]

Bone age assessment is a common diagnostic method used for abnormal growth and development in children. Despite recent significant advancements in convolutional neural network (CNN)-based intelligent bone age assessment in children, there remains room for improvement in assessment accuracy. Studies have indicated that a dual-path network (DPN) can incorporate different features of a CNN and improve the potential of the model to extract critical features compared to a single structural CNN. Attention mechanisms can also contribute to the enhanced ability of the model to extract channel and spatial features. Therefore, we propose a dual attention dual-path network (DADPN) to improve the accuracy of pediatric bone age assessment. DPN serves as a backbone network in DADPN by incorporating residual and dense connections. DPN was modified using two different attention mechanisms while containing gender information to compensate for physiological differences in bone age between males and females. Experiments were performed using this method with the RSNA Pediatric Bone Age Challenge dataset and compared to nine representative bone age assessment methods. This method achieved an optimal assessment accuracy with a mean absolute error (MAE) of 4.76 months. The experimental results suggest that the DADPN can extract the effective features of pediatric hand bone X-ray images and improve bone age assessment accuracy more than other deep learning methods. [ABSTRACT FROM AUTHOR]

Accurate power load forecasting is crucial for the sustainable operation of smart grids. However, the complexity and uncertainty of load, along with the large-scale and high-dimensional energy information, present challenges in handling intricate dynamic features and long-term dependencies. This paper proposes a computational approach to address these challenges in short-term power load forecasting and energy information management, with the goal of accurately predicting future load demand. The study introduces a hybrid method that combines multiple deep learning models, the Gated Recurrent Unit (GRU) is employed to capture long-term dependencies in time series data, while the Temporal Convolutional Network (TCN) efficiently learns patterns and features in load data. Additionally, the attention mechanism is incorporated to automatically focus on the input components most relevant to the load prediction task, further enhancing model performance. According to the experimental evaluation conducted on four public datasets, including GEFCom2014, the proposed algorithm outperforms the baseline models on various metrics such as prediction accuracy, efficiency, and stability. Notably, on the GEFCom2014 dataset, FLOP is reduced by over 48.8%, inference time is shortened by more than 46.7%, and MAPE is improved by 39%. The proposed method significantly enhances the reliability, stability, and cost-effectiveness of smart grids, which facilitates risk assessment optimization and operational planning under the context of information management for smart grid systems. [ABSTRACT FROM AUTHOR]

Determining the fundamental characteristics that define a face as "feminine" or "masculine" has long fascinated anatomists and plastic surgeons, particularly those involved in aesthetic and gender-affirming surgery. Previous studies in this area have relied on manual measurements, comparative anatomy, and heuristic landmark-based feature extraction. In this study, we collected retrospectively at Cedars Sinai Medical Center (CSMC) a dataset of 98 skull samples, which is the first dataset of this kind of 3D medical imaging. We then evaluated the accuracy of multiple deep learning neural network architectures on sex classification with this dataset. Specifically, we evaluated methods representing three different 3D data modeling approaches: Resnet3D, PointNet++, and MeshNet. Despite the limited number of imaging samples, our testing results show that all three approaches achieve AUC scores above 0.9 after convergence. PointNet++ exhibits the highest accuracy, while MeshNet has the lowest. Our findings suggest that accuracy is not solely dependent on the sparsity of data representation but also on the architecture design, with MeshNet's lower accuracy likely due to the lack of a hierarchical structure for progressive data abstraction. Furthermore, we studied a problem related to sex determination, which is the analysis of the various morphological features that affect sex classification. We proposed and developed a new method based on morphological gradients to visualize features that influence model decision making. The method based on morphological gradients is an alternative to the standard saliency map, and the new method provides better visualization of feature importance. Our study is the first to develop and evaluate deep learning models for analyzing 3D facial skull images to identify imaging feature differences between individuals assigned male or female at birth. These findings may be useful for planning and evaluating craniofacial surgery, particularly gender-affirming procedures, such as facial feminization surgery. [ABSTRACT FROM AUTHOR]

Molecular representation is critical in chemical machine learning. It governs the complexity of model development and the fulfillment of training data to avoid either over- or under-fitting. As electronic structures and associated attributes are the root cause for molecular interactions and their manifested properties, we have sought to examine the local electron information on a molecular manifold to understand and predict molecular interactions. Our efforts led to the development of a lower-dimensional representation of a molecular manifold, Manifold Embedding of Molecular Surface (MEMS), to embody surface electronic quantities. By treating a molecular surface as a manifold and computing its embeddings, the embedded electronic attributes retain the chemical intuition of molecular interactions. MEMS can be further featurized as input for chemical learning. Our solubility prediction with MEMS demonstrated the feasibility of both shallow and deep learning by neural networks, suggesting that MEMS is expressive and robust against dimensionality reduction. In machine learning of molecular properties, adequate molecular representation is crucial, as minor structural changes result in significant differences in the activity of interest. Here, the authors use a lower-dimensional representation of a molecular manifold to embed electronic attributes and retain the chemical intuition of molecular interactions, invariant to translational and rational degrees of freedom of the molecular surface, to ease model complexities and facilitate generalization even with relatively small datasets. [ABSTRACT FROM AUTHOR]

In additive manufacturing (AM), process defects such as keyhole pores are difficult to anticipate, affecting the quality and integrity of the AM-produced materials. Hence, considerable efforts have aimed to predict these process defects by training machine learning (ML) models using passive measurements such as acoustic emissions. This work considered a dataset in which keyhole pores of a laser powder bed fusion (LPBF) experiment were identified using X-ray radiography and then registered both in space and time to acoustic measurements recorded during the LPBF experiment. Due to AM's intrinsic process controls, where a pore-forming event is relatively rare, the acoustic datasets collected during monitoring include more non-pores than pores. In other words, the dataset for ML model development is imbalanced. Moreover, this imbalanced and sparse data phenomenon remains ubiquitous across many AM monitoring schemes since training data is nontrivial to collect. Hence, we propose a machine learning approach to improve this dataset imbalance and enhance the prediction accuracy of pore-labeled data. Specifically, we investigate how data augmentation helps predict pores and non-pores better. This imbalance is improved using recent advances in data augmentation called Mixup, a weak-supervised learning method. Convolutional neural networks (CNNs) are trained on original and augmented datasets, and an appreciable increase in performance is reported when testing on five different experimental trials. When ML models are trained on original and augmented datasets, they achieve an accuracy of 95% and 99% on test datasets, respectively. We also provide information on how dataset size affects model performance. Lastly, we investigate the optimal Mixup parameters for augmentation in the context of CNN performance. [ABSTRACT FROM AUTHOR]

Background: Continuous monitoring of mitral annular plane systolic excursion (MAPSE) using transesophageal echocardiography (TEE) may improve the evaluation of left ventricular (LV) function in postoperative intensive care patients. We aimed to assess the utility of continuous monitoring of LV function using TEE and artificial intelligence (autoMAPSE) in postoperative intensive care patients. Methods: In this prospective observational study, we monitored 50 postoperative intensive care patients for 120 min immediately after cardiac surgery. We recorded a set of two-chamber and four-chamber TEE images every five minutes. We defined monitoring feasibility as how often the same wall from the same patient could be reassessed, and categorized monitoring feasibility as excellent if the same LV wall could be reassessed in ≥ 90% of the total recordings. To compare autoMAPSE with manual measurements, we rapidly recorded three sets of repeated images to assess precision (least significant change), bias, and limits of agreement (LOA). To assess the ability to identify changes (trending ability), we compared changes in autoMAPSE with the changes in manual measurements in images obtained during the initiation of cardiopulmonary bypass as well as before and after surgery. Results: Monitoring feasibility was excellent in most patients (88%). Compared with manual measurements, autoMAPSE was more precise (least significant change 2.2 vs 3.1 mm, P < 0.001), had low bias (0.4 mm), and acceptable agreement (LOA − 2.7 to 3.5 mm). AutoMAPSE had excellent trending ability, as its measurements changed in the same direction as manual measurements (concordance rate 96%). Conclusion: Continuous monitoring of LV function was feasible using autoMAPSE. Compared with manual measurements, autoMAPSE had excellent trending ability, low bias, acceptable agreement, and was more precise. [ABSTRACT FROM AUTHOR]

The Drosophila model is pivotal in deciphering the pathophysiological underpinnings of various human ailments, notably aging and cardiovascular diseases. Cutting-edge imaging techniques and physiology yield vast high-resolution videos, demanding advanced analysis methods. Our platform leverages deep learning to segment optical microscopy images of Drosophila hearts, enabling the quantification of cardiac parameters in aging and dilated cardiomyopathy (DCM). Validation using experimental datasets confirms the efficacy of our aging model. We employ two innovative approaches deep-learning video classification and machine-learning based on cardiac parameters to predict fly aging, achieving accuracies of 83.3% (AUC 0.90) and 79.1%, (AUC 0.87) respectively. Moreover, we extend our deep-learning methodology to assess cardiac dysfunction associated with the knock-down of oxoglutarate dehydrogenase (OGDH), revealing its potential in studying DCM. This versatile approach promises accelerated cardiac assays for modeling various human diseases in Drosophila and holds promise for application in animal and human cardiac physiology under diverse conditions. A novel deep learning method is developed to analyse Drosophila hearts which is able to predict ageing and study cardiac dysfunction, offering potential for modeling human ailments in Drosophila [ABSTRACT FROM AUTHOR]

Aiming at the problem of image classification with insignificant morphological structural features, strong target correlation, and low signal-to-noise ratio, combined with prior feature knowledge embedding, a deep learning method based on ResNet and Radial Basis Probabilistic Neural Network (RBPNN) is proposed model. Taking ResNet50 as a visual modeling network, it uses feature pyramid and self-attention mechanism to extract appearance and semantic features of images at multiple scales, and associate and enhance local and global features. Taking into account the diversity of category features, channel cosine similarity attention and dynamic C-means clustering algorithms are used to select representative sample features in different category of sample subsets to implicitly express prior category feature knowledge, and use them as the kernel centers of radial basis probability neurons (RBPN) to realize the embedding of diverse prior feature knowledge. In the RBPNN pattern aggregation layer, the outputs of RBPN are selectively summed according to the category of the kernel center, that is, the subcategory features are combined into category features, and finally the image classification is implemented based on Softmax. The functional module of the proposed method is designed specifically for image characteristics, which can highlight the significance of local and structural features of the image, form a non-convex decision-making area, and reduce the requirements for the completeness of the sample set. Applying the proposed method to medical image classification, experiments were conducted based on the brain tumor MRI image classification public dataset and the actual cardiac ultrasound image dataset, and the accuracy rate reached 85.82% and 83.92% respectively. Compared with the three mainstream image classification models, the performance indicators of this method have been significantly improved. [ABSTRACT FROM AUTHOR]

Hepatocellular carcinoma (HCC) is a common malignancy with poor survival and requires long-term follow-up. Hence, we collected information on patients with Primary Hepatocellular Carcinoma in the United States from the Surveillance, Epidemiology, and EndResults (SEER) database. We used this information to establish a deep learning with a multilayer neural network (the NMTLR model) for predicting the survival rate of patients with Primary Hepatocellular Carcinoma. HCC patients pathologically diagnosed between January 2011 and December 2015 in the SEER (Surveillance, Epidemiology, and End Results) database of the National Cancer Institute of the United States were selected as study subjects. We utilized two deep learning-based algorithms (DeepSurv and Neural Multi-Task Logistic Regression [NMTLR]) and a machine learning-based algorithm (Random Survival Forest [RSF]) for model training. A multivariable Cox Proportional Hazards (CoxPH) model was also constructed for comparison. The dataset was randomly divided into a training set and a test set in a 7:3 ratio. The training dataset underwent hyperparameter tuning through 1000 iterations of random search and fivefold cross-validation. Model performance was assessed using the concordance index (C-index), Brier score, and Integrated Brier Score (IBS). The accuracy of predicting 1-year, 3-year, and 5-year survival rates was evaluated using Receiver Operating Characteristic (ROC) curves, calibration plots, and Area Under the Curve (AUC). The primary outcomes were the 1-year, 3-year, and 5-year overall survival rates. Models were developed using DeepSurv, NMTLR, RSF, and Cox Proportional Hazards regression. Model differentiation was evaluated using the C-index, calibration with concordance plots, and risk stratification capability with the log-rank test. The study included 2197 HCC patients, randomly divided into a training cohort (70%, n = 1537) and a testing cohort (30%, n = 660). Clinical characteristics between the two cohorts showed no significant statistical difference (p > 0.05). The deep learning models outperformed both RSF and CoxPH models, with C-indices of 0.735 (NMTLR) and 0.731 (DeepSurv) in the test dataset. The NMTLR model demonstrated enhanced accuracy and well-calibrated survival estimates, achieving an Area Under the Curve (AUC) of 0.824 for 1-year survival predictions, 0.813 for 3-year, and 0.803 for 5-year survival rates. This model's superior calibration and discriminative ability enhance its utility for clinical prognostication in Primary Hepatocellular Carcinoma. We deployed the NMTLR model as a web application for clinical practice. The NMTLR model have potential advantages over traditional linear models in prognostic assessment and treatment recommendations. This novel analytical approach may provide reliable information on individual survival and treatment recommendations for patients with primary liver cancer. [ABSTRACT FROM AUTHOR]

Social facilitation is a well-known phenomenon where the presence of organisms belonging to the same species enhances an individual organism's performance in a specific task. As far as fishes are concerned, most studies on social facilitation have been conducted in standing-water conditions. However, for riverine species, fish are most commonly located in moving waters, and the effects of hydrodynamics on social facilitation remain largely unknown. To bridge this knowledge gap, we designed and performed flume experiments where the behaviour of wild juvenile Italian riffle dace (Telestes muticellus) in varying group sizes and at different mean flow velocities, was studied. An artificial intelligence (AI) deep learning algorithm was developed and employed to track fish positions in time and subsequently assess their exploration, swimming activity, and space use. Results indicate that energy-saving strategies dictated space use in flowing waters regardless of group size. Instead, exploration and swimming activity increased by increasing group size, but the magnitude of this enhancement (which quantifies social facilitation) was modulated by flow velocity. These results have implications for how future research efforts should be designed to understand the social dynamics of riverine fish populations, which can no longer ignore the contribution of hydrodynamics. [ABSTRACT FROM AUTHOR]

Genome interpretation (GI) encompasses the computational attempts to model the relationship between genotype and phenotype with the goal of understanding how the first leads to the second. While traditional approaches have focused on sub-problems such as predicting the effect of single nucleotide variants or finding genetic associations, recent advances in neural networks (NNs) have made it possible to develop end-to-end GI models that take genomic data as input and predict phenotypes as output. However, technical and modeling issues still need to be fixed for these models to be effective, including the widespread underdetermination of genomic datasets, making them unsuitable for training large, overfitting-prone, NNs. Here we propose novel GI models to address this issue, exploring the use of two types of transfer learning approaches and proposing a novel Biologically Meaningful Sparse NN layer specifically designed for end-to-end GI. Our models predict the leaf and seed ionome in A.thaliana, obtaining comparable results to our previous over-parameterized model while reducing the number of parameters by 8.8 folds. We also investigate how the effect of population stratification influences the evaluation of the performances, highlighting how it leads to (1) an instance of the Simpson's Paradox, and (2) model generalization limitations. [ABSTRACT FROM AUTHOR]

In human–computer interaction systems, speech emotion recognition (SER) plays a crucial role because it enables computers to understand and react to users' emotions. In the past, SER has significantly emphasised acoustic properties extracted from speech signals. The use of visual signals for enhancing SER performance, however, has been made possible by recent developments in deep learning and computer vision. This work utilizes a lightweight Vision Transformer (ViT) model to propose a novel method for improving speech emotion recognition. We leverage the ViT model's capabilities to capture spatial dependencies and high-level features in images which are adequate indicators of emotional states from mel spectrogram input fed into the model. To determine the efficiency of our proposed approach, we conduct a comprehensive experiment on two benchmark speech emotion datasets, the Toronto English Speech Set (TESS) and the Berlin Emotional Database (EMODB). The results of our extensive experiment demonstrate a considerable improvement in speech emotion recognition accuracy attesting to its generalizability as it achieved 98%, 91%, and 93% (TESS-EMODB) accuracy respectively on the datasets. The outcomes of the comparative experiment show that the non-overlapping patch-based feature extraction method substantially improves the discipline of speech emotion recognition. Our research indicates the potential for integrating vision transformer models into SER systems, opening up fresh opportunities for real-world applications requiring accurate emotion recognition from speech compared with other state-of-the-art techniques. [ABSTRACT FROM AUTHOR]

The recent adoption of deep learning-based models for the processing and coding of multimedia signals has brought noticeable gains in performance, which have established deep learning-based solutions as the uncontested state-of-the-art both for computer vision tasks, targeting machine consumption, as well as, more recently, coding applications, targeting human visualization. Traditionally, applications requiring both coding and computer vision processing require first decoding the bitstream and then applying the computer vision methods to the decompressed multimedia signals. However, the adoption of deep learning-based solutions enables the use of compressed domain computer vision processing, with gains in performance and computational complexity over the decompressed domain approach. For point clouds (PCs), these gains have been demonstrated in the single available compressed domain computer vision processing solution, named Compressed Domain PC Classifier, which processes JPEG Pleno PC coding (PCC) compressed streams using a PC classifier largely compatible with the state-of-the-art spatial domain PointGrid classifier. However, the available Compressed Domain PC Classifier presents strong limitations by imposing a single, specific input size which is associated to specific JPEG Pleno PCC configurations; this limits the compression performance as these configurations are not ideal for all PCs due to their different characteristics, notably density. To overcome these limitations, this paper proposes the first Adaptive Compressed Domain PC Classifier solution which includes a novel adaptive bridge model that allows to process the JPEG Pleno PCC encoded bit streams using different coding configurations, now maximizing the compression efficiency. Experimental results show that the novel Adaptive Compressed Domain PC Classifier allows JPEG PCC to achieve better compression performance by not imposing a single, specific coding configuration for all PCs, regardless of its different characteristics. Moreover, the added adaptability power can achieve slightly better PC classification performance than the previous Compressed Domain PC Classifier and largely better PC classification performance (and lower number of weights) than the PointGrid PC classifier working in the decompressed domain. [ABSTRACT FROM AUTHOR]

In this paper, a method based on transfer learning is proposed to recover the three-dimensional shape of cultural relics faces from a single old photo. It can simultaneously reconstruct the three-dimensional facial structure and align the texture of the cultural relics with fewer training samples. The UV position map is used to represent the three-dimensional shape in space and act as the output of the network. A convolutional neural network is used to reconstruct the UV position map from a single 2D image. In the training process, the human face data is used for pre-training, and then a small amount of artifact data is used for fine-tuning. A deep learning model with strong generalization ability is trained with fewer artifact data, and a three-dimensional model of the cultural relic face can be reconstructed from a single old photograph. The methods can train more complex deep networks without a large amount of cultural relic data, and no over-fitting phenomenon occurs, which effectively solves the problem of fewer cultural relic samples. The method is verified by restoring a Chinese Terracotta Warrior with facial cracks and a Buddha with a broken nose. Other applications can be used in the fields such as texture recovery, facial feature extraction, and three-dimensional model estimation of the damaged cultural relics or sculptures in the photos. [ABSTRACT FROM AUTHOR]

Microbial necromass carbon (MNC) can reflect soil carbon (C) sequestration capacity. However, changes in the reserves of MNC in response to warming in alpine grasslands across the Tibetan Plateau are currently unclear. Based on large-scale sampling and published observations, we divided eco-clusters based on dominant phylotypes, calculated their relative abundance, and found that their averaged importance to MNC was higher than most other environmental variables. With a deep learning model based on stacked autoencoder, we proved that using eco-cluster relative abundance as the input variable of the model can accurately predict the overall distribution of MNC under current and warming conditions. It implied that warming could lead to an overall increase in the MNC in grassland topsoil across the Tibetan Plateau, with an average increase of 7.49 mg/g, a 68.3% increase. Collectively, this study concludes that alpine grassland has the tendency to increase soil C sequestration capacity on the Tibetan Plateau under future warming. A study on the carbon sequestration capacity of soil in alpine grasslands across the Tibetan Plateau implies that warming will lead to an average increase of 7.49 mg/g microbial necromass carbon in topsoil, an increase of 68.3%. [ABSTRACT FROM AUTHOR]

With the continuous progress of technology, the subject of life science plays an increasingly important role, among which the application of artificial intelligence in the medical field has attracted more and more attention. Bell facial palsy, a neurological ailment characterized by facial muscle weakness or paralysis, exerts a profound impact on patients' facial expressions and masticatory abilities, thereby inflicting considerable distress upon their overall quality of life and mental well-being. In this study, we designed a facial attribute recognition model specifically for individuals with Bell's facial palsy. The model utilizes an enhanced SSD network and scientific computing to perform a graded assessment of the patients' condition. By replacing the VGG network with a more efficient backbone, we improved the model's accuracy and significantly reduced its computational burden. The results show that the improved SSD network has an average precision of 87.9% in the classification of light, middle and severe facial palsy, and effectively performs the classification of patients with facial palsy, where scientific calculations also increase the precision of the classification. This is also one of the most significant contributions of this article, which provides intelligent means and objective data for future research on intelligent diagnosis and treatment as well as progressive rehabilitation. [ABSTRACT FROM AUTHOR]

The cast thin sections of tight oil reservoirs contain important parameters such as rock mineral composition and content, porosity, permeability and stratigraphic characteristics, which are of great significance for reservoir evaluation. The use of deep learning technology for intelligent identification of thin section images is a development trend of mineral identification. However, the difficulty of making cast thin sections, the complexity of the making process and the high cost of thin section annotation have led to a lack of cast thin section images, which cannot meet the training requirements of deep learning image recognition models. In order to increase the sample size and improve the training effect of deep learning model, we proposed a generation and annotation method of thin section images of tight oil reservoir based on deep learning, by taking Fuyu reservoir in Sanzhao Sag as the target area. Firstly, the Augmentor strategy space was used to preliminarily augment the original images while preserving the original image features to meet the requirements of the model. Secondly, the category attention mechanism was added to the original StyleGAN network to avoid the influence of the uneven number of components in thin sections on the quality of the generated images. Then, the SALM annotation module was designed to achieve semi-automatic annotation of the generated images. Finally, experiments on image sharpness, distortion, standard accuracy and annotation efficiency were designed to verify the advantages of the method in image quality and annotation efficiency. [ABSTRACT FROM AUTHOR]

Determination of protein–ligand binding affinity (PLA) is a key technological tool in hit discovery and lead optimization, which is critical to the drug development process. PLA can be determined directly by experimental methods, but it is time-consuming and costly. In recent years, deep learning has been widely applied to PLA prediction, the key of which lies in the comprehensive and accurate representation of proteins and ligands. In this study, we proposed a multi-modal deep learning model based on the early fusion strategy, called DeepLIP, to improve PLA prediction by integrating multi-level information, and further used it for virtual screening of extracellular signal-regulated protein kinase 2 (ERK2), an ideal target for cancer treatment. Experimental results from model evaluation showed that DeepLIP achieved superior performance compared to state-of-the-art methods on the widely used benchmark dataset. In addition, by combining previously developed machine learning models and molecular dynamics simulation, we screened three novel hits from a drug-like natural product library. These compounds not only had favorable physicochemical properties, but also bound stably to the target protein. We believe they have the potential to serve as starting molecules for the development of ERK2 inhibitors. [ABSTRACT FROM AUTHOR]

Proteins and nucleic-acids are essential components of living organisms that interact in critical cellular processes. Accurate prediction of nucleic acid-binding residues in proteins can contribute to a better understanding of protein function. However, the discrepancy between protein sequence information and obtained structural and functional data renders most current computational models ineffective. Therefore, it is vital to design computational models based on protein sequence information to identify nucleic acid binding sites in proteins. Here, we implement an ensemble deep learning model-based nucleic-acid-binding residues on proteins identification method, called SOFB, which characterizes protein sequences by learning the semantics of biological dynamics contexts, and then develop an ensemble deep learning-based sequence network to learn feature representation and classification by explicitly modeling dynamic semantic information. Among them, the language learning model, which is constructed from natural language to biological language, captures the underlying relationships of protein sequences, and the ensemble deep learning-based sequence network consisting of different convolutional layers together with Bi-LSTM refines various features for optimal performance. Meanwhile, to address the imbalanced issue, we adopt ensemble learning to train multiple models and then incorporate them. Our experimental results on several DNA/RNA nucleic-acid-binding residue datasets demonstrate that our proposed model outperforms other state-of-the-art methods. In addition, we conduct an interpretability analysis of the identified nucleic acid binding residue sequences based on the attention weights of the language learning model, revealing novel insights into the dynamic semantic information that supports the identified nucleic acid binding residues. SOFB is available at https://github.com/Encryptional/SOFB and https://figshare.com/articles/online_resource/SOFB_figshare_rar/25499452. We implement SOFB, an ensemble deep learning model-based nucleic-acid-binding residues on proteins identification method to characterize protein sequences by learning the semantics of biological dynamics contexts. [ABSTRACT FROM AUTHOR]

Data-driven deep learning application in earthquake engineering highlights the insufficient quantity and the imbalanced feature distribution of measured ground motions, which can be mitigated with artificial ones. Traditional ground motion generation techniques tend to extend the catalogs conditioning on existing records, while current deep learning-based methods such as generative adversarial networks (GANs) only provide limited duration or sampling rate, thus obstructing further applications. In this paper, an invertible time–frequency transformation process is employed, based on which the transformed earthquake representation is implicitly modeled by advanced GANs for high-resolution and unconditional generation. Moreover, leveraging the disentangling property of the GAN's latent space, the newly developed latent space walking method is adopted to assure the generations with controllable time–frequency features. A feature-balanced generated ground motion dataset has been constructed in combination with the proposed methods, and the application potential was demonstrated through comparative experiments of different datasets. [ABSTRACT FROM AUTHOR]

Objectives: To develop and validate a deep learning model for predicting hemorrhagic transformation after endovascular thrombectomy using dual-energy computed tomography (CT). Materials and methods: This was a retrospective study from a prospective registry of acute ischemic stroke. Patients admitted between May 2019 and February 2023 who underwent endovascular thrombectomy for acute anterior circulation occlusions were enrolled. Hemorrhagic transformation was defined using follow-up magnetic resonance imaging or CT. The deep learning model was developed using post-thrombectomy dual-energy CT to predict hemorrhagic transformation within 72 h. Temporal validation was performed with patients who were admitted after July 2022. The deep learning model's performance was compared with a logistic regression model developed from clinical variables using the area under the receiver operating characteristic curve (AUC). Results: Total of 202 patients (mean age 71.4 years ± 14.5 [standard deviation], 92 men) were included, with 109 (54.0%) patients having hemorrhagic transformation. The deep learning model performed consistently well, showing an average AUC of 0.867 (95% confidence interval [CI], 0.815–0.902) upon five-fold cross validation and AUC of 0.911 (95% CI, 0.774–1.000) with the test dataset. The clinical variable model showed an AUC of 0.775 (95% CI, 0.709–0.842) on the training dataset (p < 0.01) and AUC of 0.634 (95% CI, 0.385–0.883) on the test dataset (p = 0.06). Conclusion: A deep learning model was developed and validated for prediction of hemorrhagic transformation after endovascular thrombectomy in patients with acute stroke using dual-energy computed tomography. Clinical relevance statement: This study demonstrates that a convolutional neural network (CNN) can be utilized on dual-energy computed tomography (DECT) for the accurate prediction of hemorrhagic transformation after thrombectomy. The CNN achieves high performance without the need for region of interest drawing. Key Points: • Iodine leakage on dual-energy CT after thrombectomy may be from blood-brain barrier disruption. • A convolutional neural network on post-thrombectomy dual-energy CT enables individualized prediction of hemorrhagic transformation. • Iodine leakage is an important predictor of hemorrhagic transformation following thrombectomy for ischemic stroke. [ABSTRACT FROM AUTHOR]

Objective: This study aims to develop a weakly supervised deep learning (DL) model for vertebral-level vertebral compression fracture (VCF) classification using image-level labelled data. Methods: The training set included 815 patients with normal (n = 507, 62%) or VCFs (n = 308, 38%). Our proposed model was trained on image-level labelled data for vertebral-level classification. Another supervised DL model was trained with vertebral-level labelled data to compare the performance of the proposed model. Results: The test set included 227 patients with normal (n = 117, 52%) or VCFs (n = 110, 48%). For a fair comparison of the two models, we compared sensitivities with the same specificities of the proposed model and the vertebral-level supervised model. The specificity for overall L1–L5 performance was 0.981. The proposed model may outperform the vertebral-level supervised model with sensitivities of 0.770 vs 0.705 (p = 0.080), respectively. For vertebral-level analysis, the specificities for each L1–L5 were 0.974, 0.973, 0.970, 0.991, and 0.995, respectively. The proposed model yielded the same or better sensitivity than the vertebral-level supervised model in L1 (0.750 vs 0.694, p = 0.480), L3 (0.793 vs 0.586, p < 0.05), L4 (0.833 vs 0.667, p = 0.480), and L5 (0.600 vs 0.600, p = 1.000), respectively. The proposed model showed lower sensitivity than the vertebral-level supervised model for L2, but there was no significant difference (0.775 vs 0.825, p = 0.617). Conclusions: The proposed model may have a comparable or better performance than the supervised model in vertebral-level VCF classification. Clinical relevance statement: Vertebral-level vertebral compression fracture classification aids in devising patient-specific treatment plans by identifying the precise vertebrae affected by compression fractures. Key Points: • Our proposed weakly supervised method may have comparable or better performance than the supervised method for vertebral-level vertebral compression fracture classification. • The weakly supervised model could have classified cases with multiple vertebral compression fractures at the vertebral-level, even if the model was trained with image-level labels. • Our proposed method could help reduce radiologists' labour because it enables vertebral-level classification from image-level labels. [ABSTRACT FROM AUTHOR]