Search

Cross-project defect prediction (CPDP) plays an essential role in identifying potential defects in target projects, especially those with limited historical data, using relevant information from similar source projects. The current studies focused on three main types of software metrics for CPDP: static metrics, code-change metrics, and semantic features. However, these existing CPDP studies encounter two primary challenges: class overlap due to reduced feature dimensions and multicollinearity from integrating various software metrics. To address these challenges, we propose a CPDP model based on both reduction and hybridization techniques (RH-CPDP). The proposed model uses hybrid deep neural networks as a hybridization technique to combine the essential metrics from all metric categories, addressing the issue of class overlap to enhance prediction model efficiency. Principal component analysis (PCA) was used as a reduction method to keep the number of metrics used small, focusing on influential relationships among metrics and fault proneness and avoiding the multicollinearity problem. The experimental analysis conducted using nine open-source projects from the PROMISE dataset demonstrates that RH-CPDP surpasses current CPDP methods (TCSBoost, TPTL, DA-KTSVMO, DBN, and 3SW-MSTL) regarding area under the curve (AUC) and F1-measure. These findings highlight the effectiveness of RH-CPDP in improving the performance of CPDP techniques. • The CPDP model is proposed using reduction and hybridization techniques. • Hybrid deep neural networks are used to combine key software metrics. • Principal component analysis is used to keep the number of metrics used small. • Multiple performance indicators are used for efficiency evaluation. [ABSTRACT FROM AUTHOR]

Learning molecular representation is a crucial task in the field of drug discovery, particularly for various specific applications such as predicting molecular properties. Current methods are mainly based on deep neural network models, such as convolutional neural networks (CNN), recurrent neural networks (RNN), graph neural networks (GNN) and their mixed models. However, these neural network models mentioned above do not provide detailed explanations for the ability to learn molecular representations and why such experimental results occurred. In this paper, we aim to compare the performance of these models in predicting molecular properties based on molecular representation, and give more insight into deep neural network architectures for specific molecular tasks. Our experimental results demonstrate that the graph neural network can obtain superior performance on the regression tasks, while the mixed deep neural network models show better performance on the classification tasks. Ablation study also gives more explanation and analysis to the experimental results. [ABSTRACT FROM AUTHOR]

In the last few decades, poor prognosis of pancreatic tumor has been an issue of concern in spite of the recent advancements in the different imaging modalities. Small size, similar attenuation to normal sized pancreas or concealed position of pancreas during CT scans are the factors that leads to failure in early diagnosis of pancreatic tumor. In this research, an organized framework is proposed for monitoring, classifying and diagnosing of pancreatic tumor. The suggested model integrates the technique of Deep Neural Network (DNN) and optimistic aspects of nature-inspired algorithms; this model aims to achieve a harmonious combination of both the techniques. The proposed model examines the medical images obtained from CT scans for the presence of pancreatic tumor using SSA-ML image segmentation on CT dataset. Evaluation of suggested model in comparison to other contemporary models IDLDMS, ODL, weighted KLM, Kernel-ELM, and ELM models is also performed in reference to sensitivity, specificity, accuracy and F1 score. The classification accuracy of our model is 99.44 % which reflects its supremacy over other recent models. [ABSTRACT FROM AUTHOR]

Accurate lung cancer classification in magnetic resonance imaging (MRI) remains challenging due to the difficulty in detecting cancerous patterns. In response, this study introduces an attention-based VGG19 neural network for enhanced classification performance. Leveraging the VGG19 architecture's deep learning capabilities, our model incorporates attention mechanisms to selectively emphasize salient features during training. The attention-based approach addresses the challenge of discerning subtle patterns indicative of malignancy, significantly improving classification accuracy. We train and evaluate the model on a diverse dataset, ensuring its capacity to generalize across various patient cases. The attention mechanism proves effective in prioritizing critical regions within MRI scans, enhancing sensitivity and specificity in lung cancer detection. Additionally, we employ gradient analysis to interpret the decision-making process, providing valuable insights into influential features. Results demonstrate the proposed model's superiority over baseline approaches, showcasing its efficacy in inaccurate lung cancer classification. The attention-based VGG19 neural network not only advances classification capabilities but also offers interpretability crucial for gaining trust in automated diagnostic systems. This research contributes a robust solution to a pressing medical imaging challenge, holding promise for practical implementation in clinical settings to support radiologists in timely and accurate lung cancer diagnosis. [ABSTRACT FROM AUTHOR]

The progressive loss of motor function in the brain is a hallmark of Parkinson's disease (PD). Electroencephalogram (EEG) signals are commonly used for early diagnosis since they are associated with a brain disorder. This work aims to find a better way to represent electroencephalography (EEG) signals and enhance the classification accuracy of individuals with Parkinson's disease using EEG signals. In this paper, we present two hybrid deep neural networks (DNN) that combine convolutional neural networks with long short-term memory to diagnose Parkinson's disease using EEG signals, that is, through the establishment of parallel and series combined models. The deep CNN network is utilized to acquire the structural features of ECG signals and extract meaningful information from them, after which the signals are sent via a long short-term memory network to extract the features' context dependency. The proposed architecture was able to achieve 97.6% specificity, 97.1% sensitivity, and 98.6% accuracy for a parallel model and 99.1% specificity, 98.5% sensitivity, and 99.7% accuracy for a series model, both in 3-class classification (PD patients with medication, PD patients without medication and healthy). [ABSTRACT FROM AUTHOR]

Deep neural networks are extremely vulnerable to externalities from intentionally generated adversarial examples which are achieved by overlaying tiny noise on the clean images. However, most existing transfer-based attack methods are chosen to add perturbations on each pixel of the original image with the same weight, resulting in redundant noise in the adversarial examples, which makes them easier to be detected. Given this deliberation, a novel attentionguided sparse adversarial attack strategy with gradient dropout that can be readily incorporated with existing gradient-based methods is introduced to minimize the intensity and the scale of perturbations and ensure the effectiveness of adversarial examples at the same time. Specifically, in the gradient dropout phase, some relatively unimportant gradient information is randomly discarded to limit the intensity of the perturbation. In the attentionguided phase, the influence of each pixel on the model output is evaluated by using a soft mask-refined attention mechanism, and the perturbation of those pixels with smaller influence is limited to restrict the scale of the perturbation. After conducting thorough experiments on the NeurIPS 2017 adversarial dataset and the ILSVRC 2012 validation dataset, the proposed strategy holds the potential to significantly diminish the superfluous noise present in adversarial examples, all while keeping their attack efficacy intact. For instance, in attacks on adversarially trained models, upon the integration of the strategy, the average level of noise injected into images experiences a decline of 8. 32%. However, the average attack success rate decreases by only 0. 34%. Furthermore, the competence is possessed to substantially elevate the attack success rate by merely introducing a slight degree of perturbation. [ABSTRACT FROM AUTHOR]

In order to predict phase distributions within complex pore structures during two-phase capillary-dominated drainage, we select subsamples from computerized tomography (CT) images of rocks and simulated porous media, and develop a pore morphology-based simulator (PMS) to create a diverse dataset of phase distributions. With pixel size, interfacial tension, contact angle, and pressure as input parameters, convolutional neural network (CNN), recurrent neural network (RNN) and vision transformer (ViT) are transformed, trained and evaluated to select the optimal model for predicting phase distribution. It is found that commonly used CNN and RNN have deficiencies in capturing phase connectivity. Subsequently, we develop a higher-dimensional vision transformer (HD-ViT) that drains pores solely based on their size, regardless of their spatial location, with phase connectivity enforced as a post-processing step. This approach enables inference for images of varying sizes and resolutions with inlet-outlet setup at any coordinate directions. We demonstrate that HD-ViT maintains its effectiveness, accuracy and speed advantage on larger sandstone and carbonate images, compared with the microfluidic-based displacement experiment. In the end, we train and validate a 3D version of the model. [ABSTRACT FROM AUTHOR]

This paper presents a deep neural network incorporating visual and auditory data fusion to enhance material recognition performance. Traditional recognition techniques relying on single data modalities face accuracy and robustness limitations, especially in complex real-world environments. To address these challenges, we develop a multimodal fusion-based model. The proposed approach first extracts features from input images and sounds separately using CNNs and spectral analysis. A concatenation layer then integrates the visual and auditory features. Extensive experiments demonstrate superior material classification over uni-modal methods, with 100% test accuracy across seven material types. The multi-modal fusion model also demonstrates stronger resilience to noise and illumination variations. This research provides a valuable foundation for robust material perception in intelligent systems. [ABSTRACT FROM AUTHOR]

Longer training times pose a significant challenge in Artificial neural networks (ANNs) as it may leads to increasing the computational costs and decreasing the effectiveness of the model. Therefore, it is imperative to reduce training times in ANNs to enhance the computational efficiency. The initialization of the weights between the layers in ANN plays a vital role in reducing training times. Appropriate weight initialization can help the network converge faster during the training by providing an optimum starting point for the network. Therefore, weight initialization techniques are essential for efficient training of ANNs. This paper revisits and implements different popular weight initialization techniques in ANNs and analyzes their impact on training time. Specifically, this paper implements Gaussian-based, Kaming-based, and Xavier-based weight initiation atop a popular DNN-based network. The experiments are conducted by employing a well-known dataset. The results show that the scenario when no weight initiation is applied consumed the highest training time, whereas different weight initiation techniques contribute in reducing the training times for the network. [ABSTRACT FROM AUTHOR]

To reach the goals of low carbon emissions and a high quality of life (QOL) in Thailand, the Japan Science and Technology Agency (JST) and the Japan International Cooperation Agency (JICA) have undertaken the technical cooperation project known as the Smart Transportation Strategy for Thailand 4.0. As part of this project, an approach for estimating a QOL value from an image in a driving scene is proposed in this paper. The proposed method is based on a deep neural network. It assumes, as in the previous approach, that the QOL of a driving scene depends on the type and number of objects in the scene. Therefore, the proposed approach first applies a semantic segmentation method to an input image, and the QOL value is estimated by the deep neural network based on the MetaFormer. The loss function and optimizer are changed from the previous method to increase the accuracy of the proposed method. Smaller images than those used in the previous approach are used for faster processing. These approaches made the proposed method process more accurate and faster than the previous approach. The effectiveness of the proposed method was demonstrated by some experiments. [ABSTRACT FROM AUTHOR]

The recent advancements in wearable technology have sparked significant interest in the continuous monitoring of stress through diverse physiological indicators. Early detection of stress, facilitated by these technologies, has the potential to mitigate the adverse impacts of chronic stress and improve healthcare outcomes. This study explores stress detection using Deep Neural Networks (DNNs) trained and tested on the WESAD dataset, focusing on binary and three-level stress classification. The DNN models were optimized with the Adam and SGD optimizers across varying epochs. Results indicate that the DNN, especially when optimized with the Adam optimizer and trained with 64 epochs, achieves satisfactory results. This study highlights the efficacy of DNNs in stress assessment and underscores the importance of optimization strategies and epoch selection in achieving optimal performance. The DNN demonstrated an accuracy of 90.57% when optimized with the Adam optimizer and trained with 64 epochs for the three-level stress classification task. Moreover, its performance was nearly perfect for the binary stress classification task with accuracy of 99.7%, highlighting its robustness and efficiency across different stress level categorization scenarios. [ABSTRACT FROM AUTHOR]

This study investigates the use of neural network and their ability to predict disease progression based on clinical data and biomarkers. Using deep neural networks, a model was developed that efficiently analyzes the complex relationship between various factors and predict the probability of disease. The model was validated using retrospective analysis which indicated a good predictive ability that could be further utilized in better diagnostics and personalized treatment methods. More importantly, research detected specific pattern in the data, which enabled a more accurate prediction of disease at different stages. The study tried to improve a model by fine-tuned neural networks and tested other frameworks to gain the highest precision. This research also provides a basic for future work in directing the development of personalized therapeutic approaches based on individual patient characteristics. [ABSTRACT FROM AUTHOR]

Relating specifically to human–computer interaction (HCI), computer vision research has placed a substantial emphasis on intelligent emotion recognition in recent years. The primary emphasis lies in investigating speech aspects and bodily motions, while the knowledge of recognizing emotions from facial expressions remains relatively unexplored. Automated facial emotion detection allows a machine to assess and understand a person’s emotional state, allowing the system to predict intent by analyzing facial expressions. Therefore, this research provides a novel parameter selection strategy using swarm intelligence and a fitness function for intelligent recognition of micro emotions. This paper presents a novel method based on geometric visual representation obtained from facial landmark points. We employ the Deep Neural Networks (DNN) model to analyze the input features from the normalized angle and distance values derived from these landmarks. The results of the experiments show that Particle Swarm Optimization (PSO) worked very well by using only a few carefully chosen features. The method achieved a recognition success rate of 98.76% on the MUG dataset and 97.79% on the GEMEP datasets. [ABSTRACT FROM AUTHOR]

Background: Predicting the origin of premature ventricular contraction (PVC) from the preoperative electrocardiogram (ECG) is important for catheter ablation therapies. We propose an explainable method that localizes PVC origin based on the semantic segmentation result of a 12‐lead ECG using a deep neural network, considering suitable diagnosis support for clinical application. Methods: The deep learning‐based semantic segmentation model was trained using 265 12‐lead ECG recordings from 84 patients with frequent PVCs. The model classified each ECG sampling time into four categories: background (BG), sinus rhythm (SR), PVC originating from the left ventricular outflow tract (PVC‐L), and PVC originating from the right ventricular outflow tract (PVC‐R). Based on the ECG segmentation results, a rule‐based algorithm classified ECG recordings into three categories: PVC‐L, PVC‐R, as well as Neutral, which is a group for the recordings requiring the physician's careful assessment before separating them into PVC‐L and PVC‐R. The proposed method was evaluated with a public dataset which was used in previous research. Results: The evaluation of the proposed method achieved neutral rate, accuracy, sensitivity, specificity, F1‐score, and area under the curve of 0.098, 0.932, 0.963, 0.882, 0.945, and 0.852 on a private dataset, and 0.284, 0.916, 0.912, 0.930, 0.943, and 0.848 on a public dataset, respectively. These quantitative results indicated that the proposed method outperformed almost all previous studies, although a significant number of recordings resulted in requiring the physician's assessment. Conclusions: The feasibility of explainable localization of premature ventricular contraction was demonstrated using deep learning‐based semantic segmentation of 12‐lead ECG. Clinical trial registration: M26‐148‐8. [ABSTRACT FROM AUTHOR]

The image structuring process was mainly divided into three stages: model training, model prediction, and report structuring. In the report structure stage, based on the feature annotation sequence, this paper associated the text sequence with the corresponding table structure and stored the text sequence in the corresponding database in the background. In dataset 1, the accuracy rate of removing visual information submodel was 30 %, and that of removing semantic information submodel was 50 %. The scheme proposed in this paper was to better perform automatic image annotation and meet the requirements of image annotation in the era of Big Data. [ABSTRACT FROM AUTHOR]

With the development of Internet and AI technology, university education is facing more and more challenges and opportunities. In this paper, an intelligent learning solution based on deep neural network (DNN) and WeChat applet is proposed to solve the problems of insufficient information acquisition and difficult to meet the personalized learning needs in the learning process of college professional basic courses. To start, we construct a comprehensive knowledge map for foundational university courses across various majors utilizing deep learning techniques. This map encompasses key concepts and interconnections spanning disciplines such as mathematics, physics, and computer science. Subsequently, we devise and deploy an intelligent recommendation algorithm leveraging DNN architecture. Drawing insights from user historical data and the intricate subject interrelations within the knowledge map, our algorithm tailors personalized recommendations for learning resources. Integrating this intelligent recommendation system into the WeChat applet platform, we realize a dynamic resource recommendation service tailored to university students' foundational coursework. Empirical findings demonstrate the superior performance of our algorithm over traditional methods, showcasing enhanced accuracy, recall rates, and F1 values. Notably, users benefit from an improved recommendation efficacy and overall learning experience. This research offers a novel framework for advancing the personalized and intelligent facets of university education, bearing both theoretical significance and practical implications. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Harbin Institute of Technology. Social Sciences Edition / Haerbin Gongye Daxue Xuebao. Shehui Kexue Ban is the property of Harbin Institute of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In time-division duplex (TDD) distributed largescale multiple input multiple output (DM-MIMO) systems, the traditional downlink channel precoding method is used to resist inter-user interference (IUI). However, when the Channel State Information (CSI) is incomplete, the performance loss is serious, not only the bit error rate is high, but also the complexity of the traditional precoding algorithm is high. In order to solve these problems, this paper proposes an adaptive precoding framework based on deep learning (DL) for joint training and split application deployment. First, we train a channel emulator deep neural network (CE-DNN) to learn and simulate the transmission process of the wireless communication channel. Then, we concatenate an untrained precoding DNN (P-DNN) with a trained CE-DNN and retrain the cascaded neural network to converge. The last step is to obtain the P-DNN, namely the adaptive precoding network, by dismantling the joint trained network. Simulation results show that, when CSI is imperfect, the proposed method is compared with Tomlinson-Harashima precoding (THP) and block diagonalization (BD) precoding. The proposed method has a lower mean square error (MSE) and higher spectrum efficiency, as well as a bit error rate (BER) performance close to the THP. The source codes and the neural network codes are available on request. [ABSTRACT FROM AUTHOR]

The lung lesion segmentation from computed tomography (CT) scan image is crucial in medical imaging for identifying and treating lung disorders. To increase the precision and efficiency of lung segmentation, we can look at a new U-Net framework that makes use of attention gates and multi-head self-attention. The network may concentrate on multiple aspects of the data. The multi-head attention can gather both local as well as global information. Using numerous attention operations, the network can recognize specific aspects of lung lesions while taking surroundings into account. The attention gates work as selective amplifiers as well as suppressors in different aspects. They combine data from several scales and resolutions, emphasizing key areas and minimizing noise. By focusing on useful features, this filtering procedure improves the precision of the segmentation. This architecture uses encoders to extract low-level properties from CT images. The attention maps serve as a guide for the construction of attention gates, which judiciously select features from various U-Net levels. The lesion images are achieved through the up-sampling phases of the decoder. The lung lesion segmentation has significantly advanced due to the effective integration of multi-head self-attention and attention gates in the U-Net architecture. The proposed method AUNet-MHA with 5-fold cross-validation has achieved comparable training dice coefficient, validation dice coefficient, and test dice coefficient of 0.93, 0.92, and 0.78 respectively for lung lesion segmentation from the MedSeg CT dataset. This method captures subtle lesion patterns precisely and enhances lung lesion analysis and characterization using historical data. This method will give medical professionals more flexibility during analysis and effectively improve accuracy when diagnosing and formulating effective treatment plans. [ABSTRACT FROM AUTHOR]

Interstitial Lung Disease (ILD) is a lung illness characterized by inflammation and scarring. Identifying and categorizing ILD patterns using chest Computed Tomography (CT) images is crucial for diagnosis and treatment planning. Deep learning and computer vision advancements offer the potential for automating medical image examination, such as the transformer model, which identifies intricate dependencies and relationships in data. Chest CT scans provide valuable information for ILD pattern classification and diagnosis. The Vision Transformer (ViT) based Multi-Head Self Attention (MHSA) architecture detects local and global spatial dependencies, focusing on relevant regions and considering contextual interactions. The ViT-based model architecture aims to categorize ILD patterns using MHSA mechanisms. The proposed ViT-ILD model improves the performance by modifying hyperparameters, attention heads, and hidden units. It also utilises techniques of residual connections, layer normalization, and positional encoding for improvement. The proposed method ViT-ILD has achieved comparable training, validation and test accuracy of 100%, 98%, and 82.75% respectively for the 4-class classification with a healthy lung, hypersensitivity pneumonitis, pulmonary fibrosis, and tuberculosis from the MedGift CT dataset. It is observed that the proposed ViT-ILD model has achieved test accuracy, recall, precision, and f1-score of 82.75%, 74.15%, 100%, and 82.35%. [ABSTRACT FROM AUTHOR]

Recognition of handwritten digits has garnered significant interest among researchers in the domain of recognizing pattern. This interest stems from the recognition's relevance in various real-life applications, including reading financial checks and official documents, which has remained a persistent obstacle. To address this challenge, researchers have developed numerous algorithms focusing on recognizing handwritten digits across different human languages. This paper presents a new Kurdish Handwritten dataset, consisting of Kurdish characters, digits, texts, and symbols. The dataset consists of 1560 participants, encompassing a broad and varied group. It serves as the primary dataset for training and evaluating algorithms in Kurdish digit recognition. We used Kurdish dataset named (KurdSet) and Arabic dataset for handwritten recognition, which holds 70,000 images of Arabic digits that were written by 700 various participants. Additionally, various models are utilized in the study, including ResNet50, DenseNet121, MobileNet, and a custom CNN (convolutional neural network). Additionally, the models' effectiveness was assessed through the examination of test accuracy, which measures the percentage of correctly classified digits in the evaluation phase. ResNet50 also performs exceptionally well that achieved test accuracy 99.67%, indicating its All models exhibit good performance, DenseNet121 and the Custom CNN Model demonstrate the highest test accuracy of 99.73%, highlighting their superior performance. capabilities in capturing relevant features. Despite its accuracy, MobileNet still exhibits good recognition capability with a test accuracy 99.54%. [ABSTRACT FROM AUTHOR]

Epilepsy is a brain disorder that can affect people of all ages which is one of the most common neurological disorders affecting around 50 million individuals. Healthcare Chatbots schedule appointments, respond to patient inquiries, and offer relevant information, freeing medical staff to handle life crises. This paper implements a chatbot for epilepsy patients to increase awareness about epilepsy. Patients may easily access this AI-based healthcare chatbot whenever they have queries regarding precautions to be taken by epilepsy patients. Chatbot uses a rich set of epilepsy-related datasets collected from trusted websites which are preprocessed with NLP techniques to prepare the data for model inputs.DNN, LSTM, and RNN models are created and evaluated to produce answers to user queries related to possible precautions to avoid seizures, safety measures to be taken by patients with seizure history, etc. Evaluation of models shows that the RNN model generates more accurate answers with 94% accuracy. Finally, the epilepsy question answering Chatbot for precaution is concluded with the RNN model. [ABSTRACT FROM AUTHOR]

This manuscript investigates a fractional piecewise dengue transmission model using singular and non-singular kernels. The existence results and uniqueness of the solution are established by using the approach of fixed point and in the framework of piecewise derivative and integral. To obtain the approximate solution of the considered models we apply a piecewise numerical iteration scheme which is based on Newton interpolation polynomials. Furthermore, the numerical scheme for piecewise derivatives encompasses singular and non-singular kernels. This study aims to enhance our understanding of dengue internal transmission dynamics by using a novel piecewise derivative approach that considers both singular and non-singular kernels. This work contributes to clarifying the concept of piecewise derivatives and their significance in understanding crossover dynamics. Moreover, a deep neural network approach is employed with high accuracy in training, testing, and validation of data to investigate the specified disease problem. This methodology is employed to thoroughly investigate the intricacies of the specified disease problem. [ABSTRACT FROM AUTHOR]

Time-series forecasting is extensively used in the actual world. Recent research has shown that Transformers with a self-attention mechanism at their core exhibit better performance when dealing with such problems. However, most of the existing Transformer models used for time series prediction use the traditional encoder-decoder architecture, which is complex and leads to low model processing efficiency, thus limiting the ability to mine deep time dependencies by increasing model depth. Secondly, the secondary computational complexity of the self-attention mechanism also increases computational overhead and reduces processing efficiency. To address these issues, the paper designs an efficient multi-layer attention-based time-series forecasting model. This model has the following characteristics: (i) It abandons the traditional encoder-decoder based Transformer architecture and constructs a time series prediction model based on multi-layer attention mechanism, improving the model's ability to mine deep time dependencies. (ii) A cross attention module based on cross attention mechanism was designed to enhance information exchange between historical and predictive sequences. (iii) Applying a recently proposed sparse attention mechanism to our model reduces computational overhead and improves processing efficiency. Experiments on multiple datasets have shown that our model can significantly increase the performance of current advanced Transformer methods in time series forecasting, including LogTrans, Reformer, and Informer. [ABSTRACT FROM AUTHOR]

Heart disease is a leading cause of mortality and illness worldwide. Heart disease identification and prediction may considerably improve patient outcomes. We use deep neural networks (DNNs) and heart rate variability (HRV) data to construct a deep learning strategy for diagnosing cardiovascular abnormalities in diabetic men. The non-invasive HRV test shows how the autonomic nervous system affects heart function. It show promise for diagnosing heart dysfunction. DNNs, noted for their ability to interpret complex data patterns, are useful for prediction and diagnosis. Our unique system, DNHRV (Deep Neural Network with HRV Features), integrates two networks using DNN and DCNN methods (Deep Convolutional Neural Network). Our DNN analyses clinical risk variables using powerful deep learning architecture, while the DCNN trains. We integrate HRV signals, medical pictures, and other clinical parameters with deep neural network computing power in the suggested technique (DNNs). This multimodal technique gives us a complete picture of each patient's cardiovascular health by utilising physiological and imaging-based indicators. Our DNHRV model outperformed earlier models in accuracy, precision, F1-score, and other parameters. Our prediction model was evaluated using SHAREEDB, proving its accuracy and stability. The DNHRV model exceeds state-of-the-art CVD prediction methods by a large margin, with 98.8% accuracy, according to extensive SHAREEDB dataset tests. By highlighting CVD predicting data points, the suggested technique increased interpretability and accuracy. [ABSTRACT FROM AUTHOR]

Pneumonia is a common lung disease that is more prone to affect the elderly and those with weaker respiratory systems. However, hospital medical resources are limited, and sometimes the workload of physicians is too high, which can affect their judgment. Therefore, a good medical assistance system is of great significance for improving the quality of medical care. This study proposed an integrated system by combining transfer learning and gradient-weighted class activation mapping (Grad-CAM). Pneumonia is a common lung disease that is generally diagnosed usingX-rays.However, in areaswith limited medical resources, a shortage of medical personnel may result in delayed diagnosis and treatment during the critical period. Additionally, overworked physicians may make diagnostic errors. Therefore, having an X-ray pneumonia diagnosis assistance system is a significant tool for improving the quality of medical care. The result indicates that the best results were obtained by a ResNet50 pretrained model combined with a fully connected classification layer. A retraining procedure was designed to improve accuracy by using gradient-weighted class activation mapping (Grad-CAM), which detects the misclassified images and adds weights to them. In the evaluation tests, the final combined model is named Grad-CAM Based Pneumonia Network (GCPNet) out performed its counterparts in terms of accuracy, precision, and F1 score and reached 97.2% accuracy. An integrated system is proposed to increase model performance where Grad-CAM and transfer learning are combined. Grad-CAM is used to generate the heatmap, which shows the region that the model is focusing on. The outcomes of this research can aid in diagnosing pneumonia symptoms, as themodel can accurately classify chest X-ray images, and the heatmap can assist doctors in observing the crucial areas. [ABSTRACT FROM AUTHOR]

Rockburst is a phenomenon inwhich free surfaces are formed during excavation, which subsequently causes the sudden release of energy in the construction of mines and tunnels. Light rockburst only peels off rock slices without ejection, while severe rockburst causes casualties and property loss. The frequency and degree of rockburst damage increases with the excavation depth. Moreover, rockburst is the leading engineering geological hazard in the excavation process, and thus the prediction of its intensity grade is of great significance to the development of geotechnical engineering. Therefore, the prediction of rockburst intensity grade is one problem that needs to be solved urgently. By comprehensively considering the occurrence mechanism of rockburst, this paper selects the stress index (σθ/σc), brittleness index (σc/σt), and rock elastic energy index (Wet) as the rockburst evaluation indexes through the Spearman coefficient method. This overcomes the low accuracy problem of a single evaluation index prediction method. Following this, the BGD-MSR-DNN rockburst intensity grade prediction model based on batch gradient descent and a multi-scale residual deep neural network is proposed. The batch gradient descent (BGD) module is used to replace the gradient descent algorithm, which effectively improves the efficiency of the network and reduces the model training time. Moreover, the multi-scale residual (MSR) module solves the problem of network degradation when there are too many hidden layers of the deep neural network (DNN), thus improving the model prediction accuracy. The experimental results reveal the BGDMSR-DNNmodel accuracy to reach 97.1%, outperforming other comparable models. Finally, actual projects such as Qinling Tunnel and Daxiangling Tunnel, reached an accuracy of 100%. The model can be applied in mines and tunnel engineering to realize the accurate and rapid prediction of rockburst intensity grade. [ABSTRACT FROM AUTHOR]

Agriculture plays an important role in the economy of any country. Approximately half of the population of developing countries is directly or indirectly connected to the agriculture field. Many farmers do not choose the right crop for cultivation depending on their soil type, crop type, and climatic requirements like rainfall. This wrong decision of crop selection directly affects the production of the crops which leads to yield and economic loss in the country. Many parameters should be observed such as soil characteristics, type of crop, and environmental factors for the cultivation of the right crop. Manual decision-making is time-taking and requires extensive experience. Therefore, there should be an automated system for the right crop recommendation to reduce human efforts and loss. An automated crop recommender system should take these parameters as input and suggest the farmer's right crop. Therefore, in this paper, a long short-termmemory Network with an attention block has been proposed. The proposed model contains 27 layers, the first of which is a feature input layer. There exist 25 hidden layers between them, and an output layer completes the structure. Through these levels, the proposed model enables a successful recommendation of the crop. Additionally, the dropout layer's regularization properties aids in reduction of overfitting of the model. In this paper, a customized novel long short-term memory (LSTM) model is proposed with a residual attention block that recommends the right crop to farmers. Evaluation metrics used for the proposed model include f1-score, recall, precision, and accuracy attaining values as 95.69%, 96.56%, 96.9%, and 97.26% respectively. [ABSTRACT FROM AUTHOR]

Atrial fibrillation (AF) is one of the most common cardiac arrhythmias that affects millions of people each year worldwide and it is closely linked to increased risk of cardiovas- cular diseases such as stroke and heart failure. Machine learning methods have shown promising results in evaluating the risk of developing atrial fibrillation from the electrocardiogram. We aim to develop and evaluate one such algorithm on a large CODE dataset collected in Brazil. We used the CODE cohort to develop and test a model for AF risk prediction for individual patients from the raw ECG recordings without the use of additional digital biomarkers. The cohort is a collection of ECG recordings and annotations by the Telehealth Network of Minas Gerais, in Brazil. A convolutional neural network based on a residual network architecture was implemented to produce class probabilities for the classification of AF. The probabilities were used to develop a Cox proportional hazards model and a Kaplan-Meier model to carry out survival analysis. Hence, our model is able to perform risk prediction for the development of AF in patients without the condition. The deep neural network model identified patients without indication of AF in the presented ECG but who will develop AF in the future with an AUC score of 0. 845. From our survival model, we obtain that patients in the high-risk group (i.e. with the probability of a future AF case being >0.7) are 50% more likely to develop AF within 40 weeks, while patients belonging to the minimal-risk group (i.e. with the probability of a future AF case being less than or equal to 0.1) have >85% chance of remaining AF free up until after seven years. We developed and validated a model for AF risk prediction. If applied in clinical practice, the model possesses the potential of providing valuable and useful information in decision- making and patient management processes. [ABSTRACT FROM AUTHOR]

Stress is one of the main factors that affect the lives of many individuals' lives. Stressful work situations cause significant levels of stress in many employees. Extreme stress can cause significant problems at work and put workers' productivity, health, and safety at risk. Therefore, it is crucial to inform the person about his or her risky lifestyle and forewarn him or her before a serious issue arises. This study offers a novel optimized deep neural network technique, ABC-ABO-CNN, to forecast worker stress. Individuals are significantly impacted by stress, especially in high-stress job contexts. Identifying and measuring employee stress levels is critical to maintaining productivity, health, and safety. Traditional methods like self-assessment questionnaires have limitations, leading to biased results. The suggested approach uses a Convolutional Neural Network (CNN) in conjunction with the African buffalo and Artificial Bee Colony optimization algorithms to overcome these drawbacks. Physiological information such as heart rate, BMI, pressure, glucose level, and skin reaction is used to detect stress. The next step is to categorize stress levels using sentiment analysis. When the suggested mechanism's performance is compared to that of traditional methods, it is shown to have greater accuracy, stress detection rate, and precision values. [ABSTRACT FROM AUTHOR]

Compared to typical scanners, handheld cameras offer convenient, flexible, portable, and noncontact image capture, which enables many new applications and breathes new life into existing ones, but cameracaptured documents may suffer from distortions caused by a nonplanar document shape and perspective projection, which lead to the failure of current optical character recognition (OCR) technologies. This paper presents a new CNN model for script identification from cameracaptured Indian multilingual document images. To evaluate the performance of the proposed model 9 regional languages, one national language and one international Roman languages are considered. Two languages, Hindi national language, and Roman English language are taken as the common languages with regional language for the study. The proposed method is applied on Bi-script, Tri-script, and Multiscript combinations. The average recognition accuracy for three script combinations is 92.92%, for bi-script 91.33%, and for tri-script 87.33%. is achieved. The proposed method is the unified approach used for identifying the script from bi-script, tri-script and multi-script cameracaptured document images and is the novelty of this paper. The proposed model is compared with the Alexnet pretrained CNN model, and it achieved the highest recognition accuracy. [ABSTRACT FROM AUTHOR]

Sports injury risk prediction plays an important role in evaluating the stability of the body in the process of sports. Based on deep neural network and semi-supervised clustering theory, this paper constructs a sports injury risk prediction model, and designs a recognition algorithm based on feature sequence classification combined with Bayesian filtering, which solves the problems of false alarms and missed alarms caused by relying on local spatial attribute features. The FMS (functional movement screen) kit was used to test seven movements (squat, hurdle step, straight leg squat, shoulder flexibility, active straight leg lift, trunk stability push-up, and body rotation) of 29 athletes. In the simulation process, Pareto analysis method was used to evaluate the overall risk of athletes' after-school physical exercise, and the FMS score, receiver operating characteristic (ROC) curve and Youden index were calculated. The results showed that the best cut-off points of the total score of FMS for athletes as a whole, male athletes and female athletes were 15.5, 15.5 and 16.5, respectively, which effectively verified the generalization performance of the above sports prediction model. [ABSTRACT FROM AUTHOR]

Cloud computing technology provides flexible, on-demand, and completely controlled computing resources and services are highly desirable. Despite this, with its distributed and dynamic nature and shortcomings in virtualization deployment, the cloud environment is exposed to a wide variety of cyber-attacks and security difficulties. The Intrusion Detection System (IDS) is a specialized security tool that network professionals use for the safety and security of the networks against attacks launched from various sources. DDoS attacks are becoming more frequent and powerful, and their attack pathways are continually changing, which requiring the development of new detection methods. Here the purpose of the study is to improve detection accuracy. Feature Selection (FS) is critical. At the same time, the IDS's computational problem is limited by focusing on the most relevant elements, and its performance and accuracy increase. In this research work, the suggested Adaptive butterfly optimization algorithm (ABOA) framework is used to assess the effectiveness of a reduced feature subset during the feature selection phase, that was motivated by this motive Candidates. Accurate classification is not compromised by using an ABOA technique. The design of Deep Neural Networks (DNN) has simplified the categorization of network traffic into normal and DDoS threat traffic. DNN's parameters can be finetuned to detect DDoS attacks better using specially built algorithms. Reduced reconstruction error, no exploding or vanishing gradients, and reduced network are all benefits of the changes outlined in this paper. When it comes to performance criteria like accuracy, precision, recall, and F1-Score are the performance measures that show the suggested architecture outperforms the other existing approaches. Hence the proposed ABOA + DNN is an excellent method for obtaining accurate predictions, with an improved accuracy rate of 99.05% compared to other existing approaches. [ABSTRACT FROM AUTHOR]

Copyright of Wood Industry / Drvna Industrija is the property of Drvna Industrija and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Propagating uncertainties in kinetic models through turbulent combustion simulations to properly quantify the uncertainties in the simulation results remains a challenging and numerically expensive problem. Efficient approaches have been proposed for certain flames in the flamelet region by reducing their uncertainty input from a high-dimensional kinetic parameter space to a one-dimensional variable. However, this one-dimensional assumption does not apply to all flamelet regimes. In the current work, we developed a systematic approach to discover low-dimensional active subspace reductions that apply to the entire mixture fraction space of the flamelet, and that function even in cases where the uncertainty response is not uniform across the entire solution domain and the one-dimensional assumption does not apply. In doing so, we are able to achieve uncertainty quantification with a tunable tradeoff between high accuracy and low computational cost through careful selection of subspace dimensionality. We facilitated computation in this method using a specifically designed deep neural network based surrogate model to compute the temperature gradients of the flamelet profile to the kinetic parameters. We presented, as a proof-of-concept, a two-stage active subspace reduction on the kinetic parameter space of a nonpremixed methane flamelet. In doing so we demonstrated that its uncertainty response cannot be represented by a one-dimensional kinetic variable due to its uncorrelated behavior across the mixture fraction domain. We instead proposed a four-dimensional active subspace that captures 98 % of the uncertainty response in the flame profile at largely reduced computational cost compared to the full kinetic parameter space. The tunability, generality, and reduced computational cost of this method demonstrate its potential to facilitate uncertainty quantification of complex and large-scale combustion problems. [ABSTRACT FROM AUTHOR]

In the field of natural language processing (NLP), detecting emotions or sentiments can be a challenging task, and sometimes emotions can be more complex than just positive or negative. However, detecting sarcasm in textual data adds another layer of complexity. Despite this, identifying the underlying sarcasm in the text has become a recent area of interest among NLP researchers. Headlines in newspapers often use sarcasm to engage readers, but readers may have difficulty recognizing it, leading to a misinterpretation of the news and spreading misinformation. As a result, there is an urgent need for technology that can automatically identify sarcasm with high accuracy. Recent studies in this domain have revealed a need for a robust and efficient model. Deep learning approaches have proven to be effective in sarcasm detection. In this work, we propose a novel two-stage model that uses a word-embedding technique to select relevant features followed by an advanced deep-learning architecture to classify sarcasm in news headlines. Our proposed method demonstrates promising results in identifying sarcasm in text with an accuracy rate of approximately 97 %. We have fine-tuned the hyper-parameters to increase the precision level, which enhances the efficacy of our model. Our work provides a significant contribution to the field of NLP by presenting a reliable and effective model for sarcasm detection. The comparison of our model with recent advancements indicates that our approach outperforms them. By using our model, readers can avoid misinterpretations and the spreading of misinformation. Therefore, our work can have a positive impact on society, and we believe that it can inspire future research in the field of sarcasm detection. [ABSTRACT FROM AUTHOR]

To understand the growth of lactic acid bacteria (LAB), Limosilactobacillus fermentum, in response to medium compositions, a deep neural network (DNN) was designed using amino acids (AAs) as explanatory variables and LAB growth as the objective variable. Sixty-four different patterns of free AAs were set using an orthogonal array. The best DNN model had high accuracy with low mean square errors and predicted that Asp would affect LAB growth. Bayesian optimization (BO) using this model recommended an optimal growth media comprising maximum amounts of Asn, Asp, Lys, Thr, and Tyr and minimum amounts of Gln, Pro, and Ser. Furthermore, this proposed media was empirically validated to promote LAB growth. The absence of Gln, Ser, and Pro indicates that the different growth trends among the DNN–BO-optimized media were likely caused by the interactions among the AAs and the other components. Artificial intelligence-assisted medium optimization provided an optimal medium for Limosilactobacillus fermentum. [ABSTRACT FROM AUTHOR]

This paper presents a method for optimizing wireless networks by adjusting cell parameters that affect both the performance of the cell being optimized and the surrounding cells. The method uses multiple reinforcement learning agents that share a common policy and take into account information from neighboring cells to determine the state and reward. In order to avoid impairing network performance during the initial stages of learning, agents are pre-trained in an earlier phase of offline learning. During this phase, an initial policy is obtained using feedback from a static network simulator and considering a wide variety of scenarios. Finally, agents can intelligently tune the cell parameters of a test network by suggesting small incremental changes, slowly guiding the network toward an optimal configuration. The agents propose optimal changes using the experience gained with the simulator in the pre-training phase, but they can also continue to learn from current network readings after each change. The results show how the proposed approach significantly improves the performance gains already provided by expert system-based methods when applied to remote antenna tilt optimization. The significant gains of this approach have truly been observed when compared with a similar method in which the state and reward do not incorporate information from neighboring cells. [ABSTRACT FROM AUTHOR]

In the field of electronic warfare (EW), one of the crucial roles of electronic intelligence is the identification of radar signals. In an operational environment, it is very essential to identify radar emitters whether friend or foe so that appropriate radar countermeasures can be taken against them. With the electromagnetic environment becoming increasingly complex and the diversity of signal features, radar emitter identification with high recognition accuracy has become a significantly challenging task. Traditional radar identification methods have shown some limitations in this complex electromagnetic scenario. Several radar classification and identification methods based on artificial neural networks have emerged with the emergence of artificial neural networks, notably deep learning approaches. Machine learning and deep learning algorithms are now frequently utilized to extract various types of information from radar signals more accurately and robustly. This paper illustrates the use of Deep Neural Networks (DNN) in radar applications for emitter classification and identification. Since deep learning approaches are capable of accurately classifying complicated patterns in radar signals, they have demonstrated significant promise for identifying radar emitters. By offering a thorough literature analysis of deep learning-based methodologies, the study intends to assist researchers and practitioners in better understanding the application of deep learning techniques to challenges related to the classification and identification of radar emitters. The study demonstrates that DNN can be used successfully in applications for radar classification and identification. [ABSTRACT FROM AUTHOR]

A deep learning-based method for predicting numerous time series is suggested in this study. For example, this paper contrasts the effectiveness of DeepAR, N- Beats, Transformer, Temporal Fusion Transformer, and other deep learning models. Our test findings demonstrate that deep learning models can anticipate with excellent accuracy across a variety of time series datasets. Moreover, our results indicate that the ideal input representation differs among datasets and that it plays a significant impact in predicting accuracy. Overall, our work gives practical advice for choosing the appropriate model and input representation, as well as insights into the use of deep learning models for predicting numerous time series. [ABSTRACT FROM AUTHOR]

Fine-grained cross modal image and text retrieval is one of the hot topics in recent years, with the goal of achieving efficient and accurate image and text retrieval in different modal datasets. In this paper, a new type of cross-modal graphic retrieval method based on deep adversarial hashing is used to learn the feature representation of images and text through deep network, and realize the feature hash representation through adversarial training. The adversary network is introduced into the network structure. By maximizing the mutual information of hash coding and minimizing the anti classification loss of hash coding, the feature representation of image and text is highly consistent and separable in the hash coding space. Experiments have proved that compared with the existing algorithm, this algorithm can be retrieved better. [ABSTRACT FROM AUTHOR]

Nowadays, the quality of air in the environment has been impacted by the industry. It is important to make sure our ambient air especially in an indoor environment is clean from contaminating particles or harmful gases. Therefore, the air quality inside the indoor environment should be monitored regularly. One of the major problems, when a particular environment has been contaminated by harmful gases, is finding the source of the emission. If the indoor environment has been contaminated by a harmful source it should be instantly localized and eliminated to prevent severe casualties. In this paper, we propose the utilization of synthetic data generated by the Computational Fluid Dynamic (CFD) approach to train the Deep Neural Network (DNN) model called CFD-DNN to perform gas source localization in an indoor environment. The model is capable to locate the contaminated source within a small area of an indoor environment. A total of 361 datasets with different locations of contaminated source release have been obtained using the CFD approach. The obtained dataset was divided into training and testing datasets. The training dataset was used for the model training process while the testing dataset is fed into the model to test model reliability to predict the gas source location. The Euclidian distance equation was used to measure the distance error between the actual and predicted location of the source. The result shows that the model is capable to locate the gas source within a minimum and maximum error of 0.03m to 0.46m respectively. [ABSTRACT FROM AUTHOR]

One of JST and JICA's technical cooperation projects, "Smart Transportation Strategy to Realize Thailand 4.0," aims to reduce traffic congestion to simultaneously achieve low carbon emissions and improve citizens' quality of Life (QoL). To achieve this goal, it is necessary to develop an AI-based system that allows users to select a combination of transportation modes that meets the various needs of people of all ages and genders, and to realize this goal, evaluation of QoL from images is required. The project suggests how to improve QoL when people moves from some point to another point and some application software to recommend the time scheduling about how to optimize the QoL in a day. It recommends the highest QoL route when we have several path candidate of roads and further some application is necessary to show how QoL dynamically changes in the route. That is scenario of the QoL project. In the previous research, a method for QoL estimation from images was proposed based on semantic segmentation and MLP (Multi Layer Perceptron) by constructing a unique questionnaire form and collecting data in a car view scene. However, the previous method estimates QoL without considering differences in the attributes of the questionnaire data, so the constructed model does not satisfy the attributes of all people. In the proposed method, multiple subjects are asked to subjectively evaluate QoL in 5 levels for the same driving image data set, and the subjects are grouped by clustering based on the evaluation results. The effectiveness of the proposed method was demonstrated by comparing the MAE (mean absolute error) of the model constructed by the proposed method with that constructed by other subjects in a 5-step cross-validation. [ABSTRACT FROM AUTHOR]

In the modern digital world, the use of animated applications has increased significantly and such applications have quietly become an integral part of life. Capturing human motions is a fundamental aspect of these applications which is performed by motion-capturing (MoCap) systems. Due to the heterogeneity of MoCap systems and marker configurations, the captured skeletons may vary in their structure, number of joints, and bone lengths. The need for motion retargeting arises to transfer motion captured from one articulated actor to another. Although motion retargeting is a long-standing issue, current methods cannot automatically execute retargeting between skeletons. This paper presents the retraining of a popular motion retargeting network with a new large-scale motion dataset. The retraining aims to demonstrate the capabilities of the motion retargeting network to handle the problem of motion retargeting between skeletons that may have different structures. The paper first, presents the motion retargeting network that employs an encoder-decoder architecture. Afterward, we re-train the motion retargeting network with a new MoCap dataset. To validate the effectiveness of the network, we visually compare the generated results with the existing retargeting model. The results show that the motion retargeting network effectively retargets the different motions. [ABSTRACT FROM AUTHOR]

Embedded systems in production equipment and Internet of Things (IoT) sensors on production lines are one of the elements that constitute an industrial cyber-physical system. In this paper, an in-depth study and analysis of the optimization of blasting parameters and prediction of vibration effects in open pit mines using deep neural network arithmetic are present. Based on the deep neural network research and analysis of the relationship between blasting parameters and rock fragmentation, a prediction model for blasting parameters and fragmentation for the East Open Pit Mine was established, and sensitivity analysis was performed on blasting parameters, and the unit consumption of explosives and the perforation rate were established. It was found that the average relative errors of both numerical simulation results and depth prediction results were no more than 10%, while the average relative errors of Sadowski's formula prediction results were more than 20%. The results show that the neural network optimized by a genetic algorithm and the numerical simulation has the highest accuracy in predicting the blasting result parameters. The research model and results obtained in this paper can be used as a reference guide for engineering practice. [ABSTRACT FROM AUTHOR]

The Coronavirus outbreak (COVID-19) is still a concern for the world according to WHO. Although this virus has been controlled, prevention efforts are still being carried out. Prevention of the virus can be done by identifying patients with symptoms such as fever, respiratory distress, and sore throat. This research aims to develop an early detection system through the recognition of symptoms of COVID-19 infection using thermal camera sensors combined with the DNN using Hyperparameter Tuning. The result is the DNN algorithm can be proposed as the right algorithm to detect someone suspected of COVID-19. [ABSTRACT FROM AUTHOR]

[Display omitted] • Inconclusive single-lead ECGs generated by smart devices can cause discomfort. • Rate of inconclusive single-lead ECG can be reduced by a DNN-based algorithm. • This technology needs to be incorporated in a simple and convenient way. Several smart devices are able to detect atrial fibrillation automatically by recording a single-lead electrocardiogram, and have created a work overload at the hospital level as a result of the need for over-reads by physicians. To compare the atrial fibrillation detection performances of the manufacturers' algorithms of five smart devices and a novel deep neural network-based algorithm. We compared the rate of inconclusive tracings and the diagnostic accuracy for the detection of atrial fibrillation between the manufacturers' algorithms and the deep neural network-based algorithm on five smart devices, using a physician-interpreted 12-lead electrocardiogram as the reference standard. Of the 117 patients (27% female, median age 65 years, atrial fibrillation present at time of recording in 30%) included in the final analysis (resulting in 585 analyzed single-lead electrocardiogram tracings), the deep neural network-based algorithm exhibited a higher conclusive rate relative to the manufacturer algorithm for all five models: 98% vs. 84% for Apple; 99% vs. 81% for Fitbit; 96% vs. 77% for AliveCor; 99% vs. 85% for Samsung; and 97% vs. 74% for Withings (P < 0.01, for each model). When applying our deep neural network-based algorithm, sensitivity and specificity to correctly identify atrial fibrillation were not significantly different for all assessed smart devices. In this clinical validation, the deep neural network-based algorithm significantly reduced the number of tracings labeled inconclusive, while demonstrating similarly high diagnostic accuracy for the detection of atrial fibrillation, thereby providing a possible solution to the data surge created by these smart devices. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNNs) have become a powerful tool for a wide range of computer vision tasks, such as image classification, object detection, and semantic segmentation. This paper provides an overview of the fundamental concepts and architectures of CNNs, highlighting recent advancements and applications. We discuss the key components of CNNs including convolutional layers, pooling layers, and activation functions. The journey of CNN from its genesis to its evolution to the one we are familiar with today is covered. Although CNN is highly capable on its own many researchers have benefitted by hybridizing CNN with quality models. Therefore, we explore various tailor-made hybrid applications of CNN that are designed to solve very specific problems. We also discuss various innovative fields of applications of CNN and discuss the wide range of fields wherein CNN is performing surprisingly well. The paper concludes with future challenges and endeavors with respect to CNN. [ABSTRACT FROM AUTHOR]

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep‐learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep‐learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep‐learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator‐scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type‐based operator‐scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling‐based operator‐scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively. [ABSTRACT FROM AUTHOR]

BACKGROUND: Noninvasive respiratory support (NRS) is increasingly used to support patients with acute respiratory failure. However, noninvasive support failure may worsen outcomes compared to primary support with invasive mechanical ventilation. Therefore, there is a need to identify patients where NRS is failing so that treatment can be reassessed and adjusted. The objective of this study was to develop and evaluate 3 recurrent neural network (RNN) models to predict NRS failure. METHODS: This was a cross-sectional observational study to evaluate the ability of deep RNN models (long short-term memory [LSTM], gated recurrent unit [GRU]), and GRU with trainable decay) to predict failure of NRS. Data were extracted from electronic health records from all adult (6 18 y) patient records requiring any type of oxygen therapy or mechanical ventilation between November 1, 2013-September 30, 2020, across 46 ICUs in the Southwest United States in a single health care network. Input variables for each model included serum chloride, creatinine, albumin, breathing frequency, heart rate, SpO2, FIO2, arterial oxygen saturation (SaO2), and 2 measurements each (point-of-care and laboratory measurement) of PaO2 and partial pressure of arterial oxygen from an arterial blood gas. RESULTS: Time series data from electronic health records were available for 22,075 subjects. The highest accuracy and area under the receiver operating characteristic curve were for the LSTM model (94.04% and 0.9636, respectively). Accurate predictions were made 12 h after ICU admission, and performance remained high well in advance of NRS failure. CONCLUSIONS: RNN models using routinely collected time series data can accurately predict NRS failure well before intubation. This lead time may provide an opportunity to intervene to optimize patient outcomes. [ABSTRACT FROM AUTHOR]