Network-based Method Research Articles

A Deep Neural Network-Based Assistive Decision Method for Financial Risk Prediction in Carbon Trading Market

The price of carbon emission rights in the market fluctuates greatly due to various factors from economy, finance, and climate. For an enterprise that needs to conduct carbon trading activities, it is extremely necessary to fully grasp the price of carbon trading in the future period. Thus in this paper, a deep neural network-based assistive decision model for financial risk prediction in carbon trading market is proposed for this purpose. Specifically, the dynamic risk spillover effects of domestic and international carbon trading markets are studied, and a frontier time-varying model is utilized to measure the risk spillover effects. Then, the deep neural network is used for quantitative research and to construct an intelligent decision scheme that outputs financial risk prediction results. Four perspectives, energy price, climate environment, carbon market price, and macroeconomics, are selected as the input to analyze the influence factors of carbon emission rights price. Finally, several linear regression models are adopted as the baseline methods for comparison, and experimental results show that the proposed method can achieve better prediction performance compared with baseline methods.

Fault Diagnosis of Airborne Electronic Equipment Based on Dynamic Bayesian Networks

With the rapid development of the aerospace industry, the structure of airborne electronic equipment has become more complex, which to some extent increases the difficulty of fault detection and maintenance of airborne electronic equipment. Traditional manual fault diagnosis methods can no longer fully meet the diagnostic needs of airborne electronic equipment. Therefore, this chapter uses dynamic Bayesian network to diagnose the faults of airborne electronic equipment. The basic idea of using a dynamic Bayesian network-based fault diagnosis method for airborne electronic devices is to mine data based on historical fault data of airborne electronic devices, and obtain fault symptoms and training data of airborne electronic devices. For non-essential fault symptoms, rough set theory was introduced to reduce their attributes and obtain the simplest attribute set, thereby simplifying the network model.

Research on neural network-based fault diagnosis and prediction method for power communication equipment

Abstract In this paper, facing the digital development of the power grid and the status quo of massive power communication equipment access and targeting the demand for highly intelligent operation and maintenance management of the power grid, combined with neural network technology, we propose an intelligent diagnosis model of power communication equipment faults. Adopting BERT as the vector embedding layer to obtain the vector sequence of fault text, designing a fault entity recognition model for power communication equipment based on BERT-BiGRU-CRF, and completing the construction of the relationship set of fault text. The proposed knowledge graph-based power communication equipment fault intelligent diagnosis model combined with the WBLA-based power communication equipment fault severity level recognition algorithm to obtain different severity fault information, from which a TFIDF-COS-based power communication equipment fault intelligent diagnosis algorithm is designed to realize intelligent diagnosis of power communication equipment faults. After testing, the TFIDF-COS algorithm can get the best optimization effect when the number of hidden layers of the selected algorithm is 1, and the initial learning rate is 0.05, and its accuracy rate can be kept above 98%. Compared with the traditional fault diagnosis system, in terms of the order of magnitude 100M, 500M, 1G, and 5G, the running time is reduced by 322s, 1874s, 4617s, and 7467s, and the accuracy rate is increased by 2.33%, 2.6%, 32.02%, and 61.4%, respectively. Therefore, this paper realizes the accurate positioning of power communication equipment faults and provides technical support for intelligent operation and maintenance of the power grid.

Self-supervised neural network-based endoscopic monocular 3D reconstruction method.

Based on deep learning, monocular visual 3D reconstruction methods have been applied in various conventional fields. In the aspect of medical endoscopic imaging, due to the difficulty in obtaining real information, self-supervised deep learning has always been a focus of research. However, current research on endoscopic 3D reconstruction is mainly conducted in laboratory environments, lacking experience in dealing with complex clinical surgical environments. In this work, we use an optical flow-based neural network to address the problem of inconsistent brightness between frames. Additionally, attention modules and inter-layer losses are introduced to tackle the complexity of endoscopic scenes in clinical surgeries. The attention mechanism allows the network to better focus on pixel texture details and depth differences, while the inter-layer losses supervise the network at different scales. We have established a complete monocular endoscopic 3D reconstruction framework and conducted quantitative experiments on a clinical dataset using the cross-correlation coefficient as a metric. Compared with other self-supervised methods, our framework can better simulate the mapping relationship between adjacent frames during endoscope motion. To validate the generalization performance of our framework, we tested the model trained on the clinical dataset on the SCARED dataset and achieved equally excellent results.

GADNN: A graph attention-based method for drug-drug association prediction considering the contribution rate of different types of drug-related features

The simultaneous use of multiple drugs, known as drug combinations, is increasingly common in the treatment of complex diseases such as cancer. However, drug-drug interactions (DDIs) caused by the simultaneous use of multiple drugs can lead to unexpected side effects and even death, resulting in high medical costs. Therefore, understanding potential DDI is an essential step in reducing the risk of adverse drug reactions before administering clinical drug compounds. Recently, many machine learning-based methods for DDI prediction have been introduced, but in these methods, no attention is paid to determining the contribution rate of each type of drug feature. Here, a Graph Attention-based Deep Neural Network method called GADNN is proposed to predict DDIs based on considering the influence or contribution rate of different drug-related features. The introduced method consists of two stages. In the first stage, four basic drug datasets, including target, enzyme, infrastructure, and pathway, are used as types of drug features. Using a graph neural network-based method, the embedding vector is generated for each drug based on each dataset separately. In the second stage, the contribution coefficient of each dataset is calculated dynamically through a new graph attention mechanism. Then, the weighted combination of the separate vectors is used to predict the probability of drug-drug interactions using a dense neural network. The experimental results confirmed that the presented innovation has been able to improve the accuracy compared to state-of-the-art methods.

P4SQA: A P4 Switch-Based QoS Assurance Mechanism for SDN

Data center networks include a wide variety of service categories, which leads to a complex network topology. As a result, the QoS of different service flows cannot be guaranteed. In this paper, we propose a QoS assurance mechanism based on P4 switches (P4SQA) to make full use of network resources. P4SQA includes two functions: in-network traffic classification and in-network queue management. To reduce the communication overhead and computational pressure on the Software Defined Network (SDN) controller, we offload part of the neural computation from the control plane to the data plane. A fully connected neural network-based traffic classification method is deployed in the programmable data plane to classify different traffic types. In addition, the optimized active queue management algorithm CoDeL is integrated into the P4 switch (PCoDeL) to prevent network congestion while guaranteeing QoS more effectively. The evaluation results show that traffic classification accuracy has improved, especially in terms of the F1_score, which reached 0.97, significantly better than other state-of-the-art methods. Moreover, compared to traditional queue management algorithms, PCoDeL can improve network throughput and reduce transmission delay, which verifies the effectiveness and superiority of P4SQA in ensuring QoS.

A fatigue crack growth prediction method on small datasets based on optimized deep neural network and Delaunay data augmentation

The neural network-based prediction method has shown great potential in the field of fatigue crack growth prediction due to its powerful nonlinear processing and generalization capabilities. However, traditional deep neural network-based methods often encounter problems such as overfitting, underfitting, and instability when predicting FCG rates on small datasets. To address these issues, this study proposes a SA-DNN prediction method for metal fatigue crack growth by introducing Delaunay data augmentation and simulated annealing algorithm into the basic DNN framework. The proposed SA-DNN model is then used to predict the fatigue crack growth process of AM60B magnesium alloy and 7055 Al alloy, and the intrinsic and extrinsic generalization abilities of the model are evaluated. In both the evaluation of intrinsic and extrinsic generalization abilities, the mean squared errors of the training, validation, and testing sets rapidly converge to the target error in various epochs. The lowest mean squared errors achieved are 1.07 e-5 and 1.241 e-5, respectively. Moreover, the results also indicate that the proposed model can accurately predict the fatigue crack growth process of metal materials with varying degrees of nonlinearity.

Hemispheric lateralization of language processing: insights from network-based symptom mapping and patient subgroups.

The hemispheric laterality of language processing has become a hot topic in modern neuroscience. Although most previous studies have reported left-lateralized language processing, other studies found it to be bilateral. A previous neurocomputational model has proposed a unified framework to explain that the above discrepancy might be from healthy and patient individuals. This model posits an initial symmetry but imbalanced capacity in language processing for healthy individuals, with this imbalance contributing to language recovery disparities following different hemispheric injuries. The present study investigated this model by analyzing the lateralization patterns of language subnetworks across multiple attributes with a group of 99 patients (compared to nonlanguage processing) and examining the lateralization patterns of language subnetworks in subgroups with damage to different hemispheres. Subnetworks were identified using a whole-brain network-based lesion-symptom mapping method, and the lateralization index was quantitatively measured. We found that all the subnetworks in language processing were left-lateralized, while subnetworks in nonlanguage processing had different lateralization patterns. Moreover, diverse hemisphere-injury subgroups exhibited distinct language recovery effects. These findings provide robust support for the proposed neurocomputational model of language processing.

A Neural Network-Based Method for Real-Time Inversion of Nonlinear Heat Transfer Problems

Inverse heat transfer problems are important in numerous scientific research and engineering applications. This paper proposes a network-based method utilizing the nonlinear autoregressive with exogenous inputs (NARX) neural network, which can achieve real-time identification of thermal boundary conditions for nonlinear transient heat transfer processes. With the introduction of the NARX neural network, the proposed method offers two key advantages: (1) The proposed method can obtain inversion results using only surface temperature time series. (2) The heat flux can be estimated even when the state equation of the system is unknown. The NARX neural network is trained using the Bayesian regularization algorithm with a dataset comprising exact surface temperature and heat flux data. The neural network takes current and historical surface temperature measurements as inputs to calculate the heat flux at the current time. The capability of the NARX method has been verified through numerical simulation experiments. Experimental results demonstrate that the NARX method has high precision, strong noise resistance, and broad applicability. The composition of the input data, the surface temperature measurement noise, and the boundary heat flux shape have been studied in detail for their impact on the inversion results. The NARX method is a highly competitive solution to inverse heat transfer problems.

A Deep Convolutional Neural Network-Based Method for Self-Piercing Rivet Joint Defect Detection

Abstract The self-pierce riveting process for alloy materials has a wide range of applications in the automotive manufacturing industry. This will not only affect the operation performance but also cause accidents in severe cases when there are defects in the riveted parts. A deep learning detection model is proposed that integrates atrous convolution and dynamic convolution to identify defects of self-piercing riveting parts efficiently to overcome the problem in quality inspection after the body self-piercing riveting process. First, a backbone network for extracting riveting defect features is constructed based on the ResNet network. Second, the center area of each riveting defect is located preferentially by the center point detection algorithm. Finally, the bounding box of riveting defects is regressed to achieve defect detection based on this central region. Among them, atrous convolution is used in the external network to increase the receptive field of the model, which combined with an active convolution so that a dynamic atrous convolution module is designed. This module is used to enhance the correlation between feature points of individual pixel in the image, which helps to identify defects with incomplete image edges and suppress background interference. Ablation experiments show that the proposed method achieves the highest accuracy of 96.3%, which is 3.9% higher than the original method. It is found that the proposed method is less affected by the background interference from the qualitative comparison. Moreover, it can also effectively identify the riveting defects on the surface of each area.

A controllable neural network-based method for optimal energy management of fuel cell hybrid electric vehicles

Neural Networks (NNs) can be used for energy management of hybrid vehicles, but they are hard to tune in inference to adapt to different driving conditions. To make the NN-based energy management strategy more flexible, this paper proposes a controllable NN for optimal energy management of fuel cell hybrid electric vehicles. Inspired by the equivalent factor in the Equivalent Consumption Minimization Strategy (ECMS), we introduce an adjustable target variable for the final state as an input to the NN-based strategy. During training, classification and regression networks with single-step and multi-step inputs are considered. An efficient shooting method and an adaptive method are then introduced to realize the precise control of the final state and online parameter adaptation. Simulations of the proposed method and the benchmarking method are carried out in different battery discharge modes. Results demonstrate that the proposed shooting neural classifier can achieve 99.7% fuel optimality of dynamic programming in a similar computational time to the shooting ECMS, and the proposed adaptive neural classifier can adapt to different driving conditions and has better fuel economy than the adaptive ECMS.

Image processing and machine learning for diagnosis and screening of craniosynostosis in children

Objectivecraniosynostosis (CSO) is a congenital disorder resulting from early closure of cranial sutures in newborns, while could cause significant cosmetic and neurodevelopmental problems. As a standard method, different craniometric indices are measured directly from child head or from their 3D CT scan of skull for diagnosis or in post-operative follow-up period. We propose a novel telehealth-compatible deep learning neural network-based method for identifying different craniometric indices in non-syndromic CSO patients 2D photographic data. Methods624 pre-operative and post-operative top-down cranial digital images of 145 craniosynostotic infants (59 sagittal, 55 metopic and 31 unicoronal synostosis) who had surgery at Mofid Children’s Hospital, Tehran, Iran were used in a deep learning neural network algorithm. Head boundary was defined by a faster region-based convolutional neural network (Faster R-CNN) and then different cranial indices (cranial index (CI), cranial vault asymmetry index (CVAI), anterior-posterior width ratio (APWR), anterior-midline width ratio (AMWR) and left–right height ratio (LRHR)) were calculated from segmented images. Accuracy, sensitivity and specificity were calculated for software versus specialist data association between cranial indices were evaluated with inter-class correlation coefficients. ResultsThe head border was segmented in the proposed images with accuracy of 88.67 ± 1.94 in comparison with standard hand made procedure with a sensitivity of 86.91 ± 3.75 and specificity of 88.60 ± 4.81. Among calculated cranial indices, significant decrease in CI value is most useful for diagnosis of sagittal synostosis (CIsagittal = 71.97 ± 4.33), significant increase in CVAI value and significant decrease in LRHR value is most appropriate for unicoronal suture synostosis diagnosis (CVAIunicoronal=6.79±3.80 and LRHRunicoronal = 0.91 ± 0.05) and significant decrease in APWR and AMWR values could be indicator of metopic synostosis (AMWRmetopic= 0.77 ± 0.04 and APWRmatopic = 0.83 ± 0.05). ConclusionDeep learning neural network algorithms could have high levels of capability in calculating cranial indices from routine 2D digital images of non-syndromic craniosynostotic children and act as a substitute for optical scanner or 3D CT-based craniometrics. This method could act as a corner stone for developing a software for a mobile platform that that would allow for screening by tele-medicine or in a primary care setting.

Intelligent mesh generation for crack simulation using graph neural networks

Mesh generation for crack simulation is often the rate-limiting step because of the rapid variations in crack shape. The classical meshing paradigm, place-nodes-and-link, relies on predefined rules and fails to generalize various crack shapes. We proposed a graph neural networks-based method for recovering the missing connection information in the crack meshes. The constrained Delaunay triangulation method created a representative training mesh dataset with different crack shapes. Comprehensive and systematic analyses compare the effectiveness and efficiency of five state-of-art GNNs under different graph representations. Our study shows that the trained GraphSAGE outperforms the state-of-art GNNs on the triangular meshing task efficiently and reveals GNNs' potential to restore the missing information between adjacency vertices or edges. This work pioneers the application of GNNs for intelligent mesh generation and paves the way for complex crack simulation.

Magnetic flux leakage defect size estimation method based on physics-informed neural network.

Magnetic flux leakage (MFL) is a magnetic method of non-destructive testing for in-pipe defect detection and sizing. Despite the fact that recent developments in machine learning have revolutionized disciplines like MFL defect size estimation, the most current methods for quantifying pipeline defects are primarily data-driven, which may violate the underlying physical knowledge. This paper proposes a physics-informed neural network-based method for MFL defect size estimation. The training process of neural network is guided by the MFL data and the physical constraints that is mathematically represented by the magnetic dipole model. We use synthetic MFL data produced by a virtual MFL testing of pipeline defects to validate the proposed method through a comparison to purely data-driven neural networks and support vector machines. The findings imply that the physics-informed strategy can both improve predictive accuracy and mitigate physical violations in MFL testing, providing us with a better knowledge of how neural networks perform in defect size estimation. This article is part of the theme issue 'Physics-informed machine learning and its structural integrity applications (Part 2)'.

Neural network models for sequence-based TCR and HLA association prediction.

T cells rely on their T cell receptors (TCRs) to discern foreign antigens presented by human leukocyte antigen (HLA) proteins. The TCRs of an individual contain a record of this individual's past immune activities, such as immune response to infections or vaccines. Mining the TCR data may recover useful information or biomarkers for immune related diseases or conditions. Some TCRs are observed only in the individuals with certain HLA alleles, and thus characterizing TCRs requires a thorough understanding of TCR-HLA associations. The extensive diversity of HLA alleles and the rareness of some HLA alleles present a formidable challenge for this task. Existing methods either treat HLA as a categorical variable or represent an HLA by its alphanumeric name, and have limited ability to generalize to the HLAs that are not seen in the training process. To address this challenge, we propose a neural network-based method named Deep learning Prediction of TCR-HLA association (DePTH) to predict TCR-HLA associations based on their amino acid sequences. We demonstrate that DePTH is capable of making reasonable predictions for TCR-HLA associations, even when neither the HLA nor the TCR have been included in the training dataset. Furthermore, we establish that DePTH can be used to quantify the functional similarities among HLA alleles, and that these HLA similarities are associated with the survival outcomes of cancer patients who received immune checkpoint blockade treatments.

High-dimensional memristive neural network and its application in commercial data encryption communication

With the development of economic globalization, demands for digital business data sharing among different subsidiary companies are constantly emerging. How to achieve secure transmission of commercial data in the Internet of Things (IoT) is a challenging task. In this paper, we propose a memristive neural network-based method for encrypted communication of commercial data. First, a high-dimensional memristive Hopfield neural network (MMHNN) model with three memristive synapses and seven neurons is proposed and its complex dynamical behaviors are deeply analyzed. The results of theoretical analysis and numerical simulations show that the proposed MMHNN can generate quasi-periodic, periodic, chaotic and hyperchaotic attractors, as well as some controllable 1-direction (1D), 2D, 3D hyperchaotic multi-scroll attractors and other types of parametric regulatory dynamics and initial value regulation dynamics. In addition, an analog equivalent circuit for the MMHNN is constructed, and the circuit simulations experimental results demonstrate the correctness and feasibility of the multi-scroll attractors phenomena. Finally, a novel MMHNN-based encryption scheme for IoT commercial data images is proposed by applying hyperchaotic spatial multi-scroll attractors, and the performance analysis shows that the proposed system achieves good encryption performances, especially in terms of keyspace, information entropy, correlation and differential attacks. The hardware experiments further verify the effectiveness and superiority of the proposed commercial data encryption scheme based on the MMHNN.

A Deep Neural Network-Based Optimal Scheduling Decision-Making Method for Microgrids

With the rapid growth in the proportion of renewable energy access and the structural complexity of distributed energy systems, traditional microgrid (MG) scheduling methods that rely on mathematical optimization models and expert experience are facing significant challenges. Therefore, it is essential to present a novel scheduling technique with high intelligence and fast decision-making capacity to realize MGs’ automatic operation and regulation. This paper proposes an optimal scheduling decision-making method for MGs based on deep neural networks (DNN). Firstly, a typical mathematical scheduling model used for MG operation is introduced, and the limitations of current methods are analyzed. Then, a two-stage optimal scheduling framework comprising day-ahead and intra-day stages is presented. The day-ahead part is solved by mixed integer linear programming (MILP), and the intra-day part uses a convolutional neural network (CNN)—bidirectional long short-term memory (Bi LSTM) for high-speed rolling decision making, with the outputs adjusted by a power correction balance algorithm. Finally, the validity of the model and algorithm of this paper are verified by arithmetic case analysis.

A Neural Network-based Method for Predicting Dose to Organs at Risk in Intensity-modulated Radiotherapy for Nasopharyngeal Carcinoma

ObjectiveA neural network method was used to establish a dose prediction model for organs at risk (OARs) during intensity-modulated radiotherapy (IMRT) for nasopharyngeal carcinoma (NPC). Materials and methodsIn total, 103 patients with NPC were randomly selected for IMRT. Suborgans were automatically generated for OARs using ring structures based on distance to the target using a MATLAB program and the corresponding volume of each suborgan was determined. The correlation between the volume of each suborgan and the dose to each OAR was analysed and neural network prediction models of the OAR dose were established using the MATLAB Neural Net Fitting application. The R-value and mean square error in the regression analysis were used to evaluate the prediction model. ResultsThe OAR dose was related to the volume of the corresponding sub-OAR. The average R-values for the normalised mean dose (Dnmean) to parallel organs and serial organs and the normalised maximum dose (Dn0) to serial organs in the training set were 0.880, 0.927 and 0.905, respectively. The mean square error for each OAR in the prediction model was low (ranging from 1.72 × 10−4 to 7.06 × 10−3). ConclusionThe neural network-based model for predicting OAR dose during IMRT for NPC is simple, reliable and worth further investigation and application.

Network-based uncertainty quantification for mathematical models in epidemiology

After the new Coronavirus disease (COVID-19) emerged in the end of January 2020 in Germany, a large number of individuals suffered from severe symptoms and eventually needed intensive care in hospitals. Due to the rapid spread of the disease, the number of deceased individuals increased as well, which is a motivation to prevent as many new infections as possible. Therefore, the knowledge about the current evolution of the virus spread is crucial to predict its future behavior and to react with suitable interventions. In this paper, the evolution of the COVID-19 pandemic in Germany is forecasted by a network-based inference method, in which the interactions of individuals are taken into account using a contact matrix. Then the results are compared to the predictions without considering a contact matrix as well as to the logistic regression, which shows the advantage of incorporating the contact matrix. Furthermore, the basic reproduction number of the pandemic in Germany using a neural network approach is estimated and used for further predictions of the evolution of COVID-19 in Germany. In order to mathematically model the different compartments of the population in the considered regions, the classical SIR model is employed. In this work, we deploy the LASSO (Least Absolute Shrinkage and Selection Operator) for the unknown parameter estimation. Furthermore, we calculate and illustrate the MAPE (Mean Absolute Percentage Error) of the estimations to show the accuracy of the predictions. The results include model parameter estimation and model validation, as well as the outbreak forecasting using network-informed algorithms. Our findings show that the network-inference based approach outperforms the logistic regression as well as the neural network approach and the SIR model calibration without a contact network. Furthermore according to the results, the network-inference based approach is particularly suitable for short- to mid-term predictions, even when there is not much information about the new disease. Moreover, the predictions based on the estimation of the reproduction number in Germany can yield more reliable results with increasing the availability of data, but could not outperform the network-inference based algorithm.

Graph neural network-based similarity relationship construction model for geospatial services

ABSTRACT During the development of service-based software systems, Geospatial Service (GS) replacement is often performed, which requires the discovery of functionally similar services in service registries to replace failed services. Compared to real-time similarity computations, direct extraction of similar services from constructed similarity relationships can yield higher replacement efficiency. However, missing and inconsistent service-registry information impedes accurate similarity relationship construction. Here, we propose a Graph Neural Network (GNN)-based model for GS Similarity Relationship construction considering service descriptions and tags, which is named GSSR-GNN. As the sparsity of the service similarity relationship graph constructed based on labeled samples limits the information propagation ability, a graph augmentation method for similarity relationship construction among second-order neighbors is proposed. Considering the differences in the semantic-information feature distributions, such as the service descriptions and tags, a feed-forward neural network-based fusion method is designed to embed them into the same vector space. Pre-trained Bidirectional Encoder Representations from Transformers (BERT) and WordNet models are introduced to enhance the service-representation expressiveness. When an enhanced service representation is input to the GNN, the similarity is calculated and the service similarity relationship is obtained. Experimental results show that the proposed model constructs service similarity relationships with high precision, thus improving the service replacement efficiency and reducing the computational cost of service registry during service replacement.