Neural Network Method Research Articles

High-Efficiency Finite Element Model Updating of Bridge Structure Using a Novel Physics-Guided Neural Network

An accurate finite element model (FEM) plays a critical role in the structural damage identification. However, due to the existence of the uncertainties, such as material properties and modeling errors, it always exists some gaps between the analytical FEM and experimental structure. While an artificial neural network (ANN)-based model updating methods have been widely adopted to narrow the gap and obtain a baseline FEM, it still faces inaccurate results and fails to meet the physical law. In this regard, the study proposes a novel physics-based loss function inspired by modal sensitivity analysis and incorporates it into the residual neural network, thereby forming a novel physics-guided neural network (PGNN) method. The mapping relationship between the input of structural responses and output of model updating variables is constrained to retain its physical meaning by guiding the training process instead of pure data association, which aims to improve the accuracy of the ANN-based method and achieve accurate and high-efficiency model updating. An experimental example of a continuous rigid frame bridge is adopted to verify the feasibility of the proposed method. Additionally, other common model updating methods, including moth-flame optimization and regularization method, are used to make a comparison. The noise-robustness of the proposed method is investigated as well. Compared to the existing method, the results illustrate that the proposed PGNN method can achieve better model updating and good noise-robustness under high uncertainties, which means the introduction of the physics-based loss function significantly enhances the parameters updating ability of the neural network. The proposed method exhibits high efficiency and promising potential for large-scale bridge structure model updating.

A Review of Hypergraph Neural Networks

In recent years, Graph Neural Networks (GNNs) have seen notable success in fields such as recommendation systems and natural language processing, largely due to the availability of vast amounts of data and powerful computational resources. GNNs are primarily designed to work with graph data that involve pairwise relationships. However, in many real-world networks, the relationships between entities are complex and go beyond simple pairwise connections, as seen in scientific collaboration networks, protein networks, and similar domains. If these complex relationships are directly represented as pairwise relationships using graph structures, it can lead to information loss. A hypergraph, as a special kind of graph-structured data, can represent higher-order relationships that cannot be fully captured by graphs, thereby addressing the limitations of graphs. In light of this, researchers have begun to focus on how to design neural networks on hypergraphs, leading to the proposal of hypergraph neural network (HGNN) models for downstream tasks. Therefore, this paper reviews the existing hypergraph neural network models. The review is conducted from two perspectives: spectral analysis methods and neural network methods on hypergraphs, discussing both unfolded and non-unfolded methods, and further subdividing them based on their algorithm characteristics and application scenarios. Subsequently, the design concepts of various algorithms are analyzed and compared, and the advantages and disadvantages of each type of algorithm are summarized based on experimental results. Finally, potential future research directions in hypergraph learning are discussed.

Respiratory disease detection in lung auscultation with convolutional neural networks and CVAE augmentation

The main purpose of this work was to investigate the possibility of detecting respiratory diseases in audio recordings of lung auscultation using modern deep learning tools, as well as to explore the possibility of using data augmentation by generating synthetic spectral representations of audio samples. The ICBHI (International Conference on Biomedical and Health Informatics) dataset was used for training, validation and augmentation. The dataset includes lung auscultations of 126 different subjects, there are a total of 920 sounds, of which 810 have signs of chronic diseases, 75 of non-chronic diseases and 35 with no pathology. The stage of data preprocessing includes discretization to 4kHz frequency, as well as filtering of frequency bands that do not carry information value for the task. In the next step, each sample was transformed into a frequency spectrum and Melspectrograms were generated. To solve the problem of class imbalance, the required number of synthetic spectrograms generated by convolutional variation autoencoders was added. At the stage of building the model, the methods of classical convolutional neural networks were used. The quality of the obtained algorithm was evaluated using a 10-fold cross-validation. Also, to assess the generalization of the proposed method, experiments were performed with the split of audio recordings into training and test sets using patient grouping. Qualitative evaluation of the model was performed using sensitivity, specificity, F1-score and Cohen’s kappa. A score of 98.45% F1-score was achieved for the 5-class classification problem which can contribute to the development of ways to synthesize and augment sensitive medical data. In addition, a cons of existing methods in the generalization of the obtained predictions were revealed, which opens the way for further research in the direction of clinical respiratory diseases detection.

The exact analytical solutions of the (2+1)-dimensional extended Korteweg–de Vries equation using bilinear neural network method and bilinear residual network method

The [Formula: see text]-dimensional extended Korteweg–de Vries (KdV) equation is predominantly utilized to elucidate the propagation of waves that exhibit both dispersive and nonlinear characteristics within the domain of nonlinear physics. This paper employs the bilinear neural network method (BNNM) to derive the exact analytical solutions of the equation. By constructing various bilinear neural network models, we obtain lump solution, breather solution and periodic interaction solution of the equation. The bilinear residual network method (BRNM) is an extension of BNNM. We apply BRNM under specific constraints to obtain breather solution of the equation, thereby offering a broader conceptual framework. Subsequently, various 3D plots, contour plots, density plots and x-curves are used to illustrate the physical properties and dynamic behaviors of these waves.

Periodic solutions, breather, lump and interaction solutions of a generalized (2+1)-dimensional Hirota bilinear equation via bilinear neural network method

Abstract The main focus of this paper is to address a generalized (2+1)-dimensional Hirota bilinear equation utilizing the bilinear neural network method. The paper presents the periodic solutions through a single-layer model of [3-4-1], followed by breather, lump and their interaction solutions by using double-layer models of [3-3-2-1] and [3-3-3-1], respectively. A significant innovation introduced in this work is the computation of periodic cross-rational solutions through the design of a novel [3-(2+2)-4-1] model, where a specific hidden layer is partitioned into two segments for subsequent operations. Three-dimensional and density figures of the solutions are given alongside an analysis of the dynamics of these solutions.

Uninvolved liver dose prediction in stereotactic body radiation therapy for liver cancer based on the neural network method.

The quality of a radiotherapy plan often depends on the knowledge and expertise of the plan designers. To predict the uninvolved liver dose in stereotactic body radiotherapy (SBRT) for liver cancer using a neural network-based method. A total of 114 SBRT plans for liver cancer were used to test the neural network method. Sub-organs of the uninvolved liver were automatically generated. Correlations between the volume of each sub-organ, uninvolved liver dose, and neural network prediction model were established using MATLAB. Of the cases, 70% were selected as the training set, 15% as the validation set, and 15% as the test set. The regression R-value and mean square error (MSE) were used to evaluate the model. The volume of the uninvolved liver was related to the volume of the corresponding sub-organs. For all sets of R-values of the prediction model, except for Dn0 which was 0.7513, all R-values of Dn10-Dn100 and Dnmean were > 0.8. The MSE of the prediction model was also low. We developed a neural network-based method to predict the uninvolved liver dose in SBRT for liver cancer. It is simple and easy to use and warrants further promotion and application.

Deep neural network-based molecular dynamics simulations for AlxGa1-xN alloys and their thermal properties

Efficient heat dissipation is crucial for the performance and lifetime of high electron mobility transistors (HEMTs). The thermal conductivity of materials and interfacial thermal conductance (ITC) play significant roles in their heat dissipation. To predict the thermal properties of AlxGa1-xN and the ITC of GaN/AlxGa1-xN in HEMTs, a dataset with first-principles accuracy was constructed using concurrent learning method and trained to obtain an interatomic potential employing deep neural networks (DNN) method. Using obtained DNN interatomic potential, equilibrium molecular dynamics simulations were employed to calculate the thermal conductivity of AlxGa1-xN, which showed excellent consistent with experimental results. Additionally, the phonon density of states of AlxGa1-xN and the ITC of GaN/AlxGa1-xN were calculated. Our study revealed a decrease in the ITC of GaN/AlxGa1-xN with increasing x, and the insertion of 1nm-thick AlN at the interface significantly reduced the ITC. This work provided a high-fidelity DNN potential for molecular dynamics simulations of AlxGa1-xN, offering valuable guidance for exploring the thermal transport of complex alloy and heterostructure.&#xD.

A graph similarity measurement algorithm combining global and local fine-grained features

Graph Similarity Computation (GSC) is not only a basic operation of graph similarity search, but also plays an important role in many application fields, including drug design, program analysis and social group identification. Since calculating the exact distance or similarity between two graphs is usually an NP-hard problem, the tradeoff between precision and speed needs to be addressed. In this paper, a pooling-based graph neural network method is proposed, which effectively integrates the coarse-grained interaction features of the two graph data and the fine -grained interaction features of the nodes between the subgraphs, and further reduce the computational cost while ensuring the accuracy. The experimental results show that the proposed method has good performance on real graph data sets. Compared with previous methods, the proposed method not only improves the accuracy but also improves the computational efficiency.

An efficient convolution neural network method for copy-move video forgery detection

Unmanned systems play a pivotal role in military surveillance, critical infrastructure protection, law enforcement, search and rescue operations, and border security, showcasing their multifaceted importance across diverse applications. Video fraud detection is integral to multimedia security, where our task involves the precise identification of modified segments within video sequences. Current approaches to video fraud detection often rely on manual feature selection and models tailored to detect specific tampering types, such as copy-move or splicing. The general representation powers of deep learning models and the connection of multiple forensic characteristics are still not fully explored. This research uses a convolutional neural network (CNN) model to identify copy-move video forgeries. Copy-move forgery is a type of video tampering whereby a portion of the video is copied and pasted somewhere different in the same video to cover an essential video characteristic. The method that is being proposed involves dividing the video into individual frames, extracting features from each frame by using a CNN model that has already been trained, and then utilizing these features to train a new CNN model that would classify each frame as either legitimate or fabricated. The proposed method effectively detects copy-move video forgery with an exceptionally high accuracy rate, exceeding current methods of accuracy and computational effort. The proposed method outperformed all other approaches on the SULFA, GRIP, and VTD datasets. The model's accuracy was 85.42 %, 86.16 %, and 81.87 %, respectively, with the shortest times recorded being 9.6 sec, 11.4 sec, and 13.7 sec, respectively. Consequently, specialists can employ the suggested method as a machine-learning instrument for detecting fake videos in real-time.

A deep equivariant neural network approach for efficient hybrid density functional calculations

Hybrid density functional calculations are essential for accurate description of electronic structure, yet their widespread use is restricted by the substantial computational cost. Here we develop DeepH-hybrid, a deep equivariant neural network method for learning the hybrid-functional Hamiltonian as a function of material structure, which circumvents the time-consuming self-consistent field iterations and enables the study of large-scale materials with hybrid-functional accuracy. Our extensive experiments demonstrate good reliability as well as effective transferability and efficiency of the method. As a notable application, DeepH-hybrid is applied to study large-supercell Moiré-twisted materials, offering the first case study on how the inclusion of exact exchange affects flat bands in magic-angle twisted bilayer graphene. The work generalizes deep-learning electronic structure methods to beyond conventional density functional theory, facilitating the development of deep-learning-based ab initio methods.

A comprehensive survey of space robotic manipulators for on-orbit servicing.

On-Orbit Servicing (OOS) robots are transforming space exploration by enabling vital maintenance and repair of spacecraft directly in space. However, achieving precise and safe manipulation in microgravity necessitates overcoming significant challenges. This survey delves into four crucial areas essential for successful OOS manipulation: object state estimation, motion planning, and feedback control. Techniques from traditional vision to advanced X-ray and neural network methods are explored for object state estimation. Strategies for fuel-optimized trajectories, docking maneuvers, and collision avoidance are examined in motion planning. The survey also explores control methods for various scenarios, including cooperative manipulation and handling uncertainties, in feedback control. Additionally, this survey examines how Machine learning techniques can further propel OOS robots towards more complex and delicate tasks in space.

Forecasting Occupancy Rate using Neural Network and Decision Tree at Hotel X

Occupancy rate is a critical factor in a hotel because it is used to measure the operational success of a hotel. The higher the hotel occupancy rate, the more successful the hotel business is in generating revenue. In the hospitality business, occupancy rates are very important to monitor and analyse as marketing strategies and pricing policies. This study compares the forecasting of occupancy rates at Hotel X using neural network method and decision tree method. The dataset used in this study is room available, room sold, and available occupancy percentage data at Hotel X from April 2018 to June 2023. The simulation was carried out by dividing the data into training data and testing data with a percentage ratio of 70:30, 75:25, 80:20, 85:15, and 90:10. The forecasting results that have been carried out using a neural network with one hidden layer have an optimal RSME result of 0.010 for split data of 70%:30% and 80%:20% while using a neural network with two hidden layers the optimal result of RSME is 0.013 for split data of 75%:25%. Forecasting results using decision tree RSME optimal results of 0.022 for split data 85%:15% and 90%:10%. From this forecasting, the most optimal results use a neural network with one hidden layer for data splits of 70%:30% and 80%:20% with RSME results of 0.010. The results of the research can be used by Hotel X as a policy determination in the next hotel management.

A Novel Wiener‐Type Dynamic Neural Network Method for Large Signal Modeling of Power Amplifiers

ABSTRACTA novel Wiener‐type dynamic neural network (Wiener‐type DNN) method for large signal modeling of power amplifiers (PAs) is proposed in this paper. In this method, the proposed model structure for the PA contains two Wiener‐type DNNs to describe the input impedance and the amplification efficiency, respectively. The Wiener‐type DNN includes a simplified linear dynamic equation module and a nonlinear static equation module. For the simplification process of the linear dynamic equation, it is proposed to process fundamental frequency components of the large signal harmonic data of the PA by vector fitting. The neural network is used to implement the static equation in the Wiener‐type DNN. The formulas for training the proposed Wiener‐type DNN model of the PA using large signal data are derived. The training algorithm of the PA model is proposed to improve the training efficiency. A Motorola MOSFET PA example and a Freescale lateral double‐diffused MOSFET PA example are present to validate the proposed Wiener‐type DNN method. For the Motorola MOSFET PA, a dynamic neural network (DNN) model and a time‐delay deep neural network (TDDNN) are established for comparison. For the Freescale PA, the DNN model is used for comparison experiment. The comparison figures show that the Wiener‐type DNN model has better convergence properties than the DNN model in the nonlinear region of the PA. The established accurate PA model is conducive to the design of subsequent communication circuits.

A semantic segmentation method integrated convolutional nonlinear spiking neural model with Transformer

Semantic segmentation is a critical task in computer vision, with significant applications in areas like autonomous driving and medical imaging. Transformer-based methods have gained considerable attention recently because of their strength in capturing global information. However, these methods often sacrifice detailed information due to the lack of mechanisms for local interactions. Similarly, convolutional neural network (CNN) methods struggle to capture global context due to the inherent limitations of convolutional kernels. To overcome these challenges, this paper introduces a novel Transformer-based semantic segmentation method called NSNPFormer, which leverages the nonlinear spiking neural P (NSNP) system—a computational model inspired by the spiking mechanisms of biological neurons. The NSNPFormer employs an encoding–decoding structure with two convolutional NSNP components and a residual connection channel. The convolutional NSNP components facilitate nonlinear local feature extraction and block-level feature fusion. Meanwhile, the residual connection channel helps prevent the loss of feature information during the decoding process. Evaluations on the ADE20K and Pascal Context datasets show that NSNPFormer achieves mIoU scores of 53.7 and 58.06, respectively, highlighting its effectiveness in semantic segmentation tasks.

Pose Estimation of a Cobot Implemented on a Small AI-Powered Computing System and a Stereo Camera for Precision Evaluation.

The precision of robotic manipulators in the industrial or medical field is very important, especially when it comes to repetitive or exhaustive tasks. Geometric deformations are the most common in this field. For this reason, new robotic vision techniques have been proposed, including 3D methods that made it possible to determine the geometric distances between the parts of a robotic manipulator. The aim of this work is to measure the angular position of a robotic arm with six degrees of freedom. For this purpose, a stereo camera and a convolutional neural network algorithm are used to reduce the degradation of precision caused by geometric errors. This method is not intended to replace encoders, but to enhance accuracy by compensating for degradation through an intelligent visual measurement system. The camera is tested and the accuracy is about one millimeter. The implementation of this method leads to better results than traditional and simple neural network methods.

Predicting SARS-CoV-2 infection among hemodialysis patients using deep neural network methods

COVID-19 has a higher rate of morbidity and mortality among dialysis patients than the general population. Identifying infected patients early with the support of predictive models helps dialysis centers implement concerted procedures (e.g., temperature screenings, universal masking, isolation treatments) to control the spread of SARS-CoV-2 and mitigate outbreaks. We collect data from multiple sources, including demographics, clinical, treatment, laboratory, vaccination, socioeconomic status, and COVID-19 surveillance. Previous early prediction models, such as logistic regression, SVM, and XGBoost, require sophisticated feature engineering and need improved prediction performance. We create deep learning models, including Recurrent Neural Networks (RNN) and Convolutional Neural Networks (CNN), to predict SARS-CoV-2 infections during incubation. Our study shows deep learning models with minimal feature engineering can identify those infected patients more accurately than previously built models. Our Long Short-Term Memory (LSTM) model consistently performed well, with an AUC exceeding 0.80, peaking at 0.91 in August 2021. The CNN model also demonstrated strong results with an AUC above 0.75. Both models outperformed previous best XGBoost models by over 0.10 in AUC. Prediction accuracy declined as the pandemic evolved, dropping to approximately 0.75 between September 2021 and January 2022. Maintaining a 20% false positive rate, our LSTM and CNN models identified 66% and 64% of positive cases among patients, significantly outperforming XGBoost models at 42%. We also identify key features for dialysis patients by calculating the gradient of the output with respect to the input features. By closely monitoring these factors, dialysis patients can receive earlier diagnoses and care, leading to less severe outcomes. Our research highlights the effectiveness of deep neural networks in analyzing longitudinal data, especially in predicting COVID-19 infections during the crucial incubation period. These deep network approaches surpass traditional methods relying on aggregated variable means, significantly improving the accurate identification of SARS-CoV-2 infections.

Research on the prediction method of wing structure noise based on the combination of conditional generative adversarial neural network and numerical methods

This paper presents a technique for predicting noise generated by airfoil structures that combines deep learning techniques with traditional numerical methods. In traditional numerical methods, accurately predicting the noise of airfoil structures requires significant computational resources, making it challenging to perform low-noise optimization design for these structures. To expedite the prediction process, this study introduces Conditional Generative Adversarial Networks (CGAN). By replacing the generator and discriminator of CGAN with traditional regression neural network models, the suitability of CGAN for regression prediction is ensured. In this study, the data computation was accelerated by expanding the kernel function in the traditional boundary element method using a Taylor series. Based on the resulting data, an alternative predictive model for wing structure noise was developed by integrating Conditional Generative Adversarial Networks (CGAN). Finally, the effectiveness and feasibility of the proposed method are demonstrated through three case studies.

Prediction of mechanical property of open-hole composite laminates using generalized regression neural network method

Prediction of mechanical property of open-hole composite laminates using generalized regression neural network method

Redundancy Design of FADS System for Complex Leading- Edge Aircraft Based on Neural Network

The flush airdata sensing system (FADS) calculates flight parameters such as angle of attack, sideslip angle, Mach number, incoming flow pressure and static pressure based on the pressure measurement of the aircraft surface. It can effectively solve the problem that the leading edge of the pitot tube protruding from the fuselage cannot adapt to the severe aerodynamic heating faced by hypersonic aircraft during the cruise phase, and at the same time meet the aircraft's demand for stealth performance. At present, there is little analysis and research on the use of neural network methods and FADS systems for complex leading edge aircraft. In view of the subsonic/transonic conditions of hypersonic aircraft returning autonomously during the landing phase, the redundancy design and verification of the head FADS system of complex leading edge aircraft are carried out considering the influence of thin leading edge and inlet components. 15 pressure measuring holes are opened in the head of the complex leading edge aircraft. Through a large number of refined numerical simulations, the pressure database of the aircraft under different incoming flow conditions is established, and the typical working conditions are verified by wind tunnel tests. For the complex leading edge aircraft, 4 sets of neural network algorithms are established based on pressure data and redundant design research is carried out, including 1 set of 9-hole algorithm and 3 sets of redundant algorithms. Among them, the 9-hole algorithm has a higher accuracy, with the solution error of the angle of attack within 0.07∘, the solution error of the sideslip angle within 0.3∘, the solution error of the Mach number within 0.0012, and the relative error of the solution of the incoming flow pressure and static pressure within 1.5%. In addition, a system solution process with a certain fault tolerance was established, which can continue to maintain the effective output of the incoming flow parameters in the event of failure of any single pressure measuring hole.

Comparison of measurement of integrated relaxation pressure by esophageal manometry with analysis of swallowing sounds with artificial intelligence in patients with achalasia.

Esophageal motility disorders are mainly evaluated with high-resolution manometry (HRM) which is a time-consuming and uncomfortable procedure with potential adverse events. Acoustic characterization of the swallowing has the potential to be an alternative noninvasive procedure. We compared the findings on HRM and swallowing sounds in 43 patients who were referred for evaluation of dysphagia. The sound analysis was done with empirical mode decomposition method and with artificial intelligence (AI) and the estimated integrated relaxation pressure (IRP) from a two-layer neural network method was compared to measured IRP on HRM. The model then was tested in five patients. IRP was estimated with high accuracy using the model developed with two-layer neural network method. The analysis of acoustic properties of swallowing has the potential to be used for evaluation of esophageal motility disorders, this needs to be further evaluated in larger studies.