Neural Network Method Research Articles

COVID-19 REINFECTION PROCESSES: MATHEMATICAL MODELING, DATA FITTING WITH TIME-VARYING PARAMETERS, AND OPTIMAL CONTROL STUDIES

In this paper, we construct a COVID-19 dynamic model that included both the initial and reinfected population compartments, and conduct a structural identifiability analysis of the model parameters to ensure the robustness of the parameter fitting results. We use some actual statistical data from North Carolina to fit the model and estimate the values of some important parameters. In order to accurately fit the parameters in the model, we improve the physics-informed neural networks (PINNs) method in this paper, so that the fitting results can be reproduced on Matlab. The results of this study show that the transmission capacity of the virus in the reinfected person is only slightly lower than that of the first infected person, and vaccination is not effective in reducing the transmission rate of the virus. The death rate of the reinfected is much higher than that of the first infected person. Finally, we conduct a cost-effectiveness study using optimal control methods and found that, while it is easier to reduce reinfection by combining multiple strategies, the most effective strategy for reducing reinfection is to increase treatment cure rates and reduce direct or indirect contact among those who have recovered.

Prediction of the hot flow behavior of AISI 321 stainless steel during dynamic recrystallization using sine hyperbolic constitutive equation and artificial neural network

The hot compression deformation behavior of AISI 321 austenitic stainless steel were studied under the strain rate and temperature ranges of 0.001–1 s−1 and 950–1200°C. The single peak hot flow curves indicated the occurrence of dynamic recrystallization in AISI 321 steel at applied deformation conditions. The microstructural observations showing the development of equaxed austenite grains further approved the results of flow curve analysis. The average size of austenite grains, when deformation is carried out at strain rate of 1 s−1, increases from 8 μm at 950°C to 61 μm at deformation temperature of 1200°C. After microstructure analysis, the flow behavior of AISI 321 steel was modeled using sine-hyperbolic constitutive equation and feed-forward back propagation artificial neural network. The mean absolute error related to the predicted flow stress values at different strains with the sine-hyperbolic constitutive equation is more than 2, and this value is less than unity for the neural network method. The obtained results efficiently indicate that the ANN model is more accurate in predicting the flow behavior of AISI 321 austenitic stainless steel when the dynamic recrystallization is the prevailing softening mechanism.

Software Reliability Prediction Based on Recurrent Neural Network and Ensemble Method

Software Reliability Prediction Based on Recurrent Neural Network and Ensemble Method

Mining and analysis of adverse event signals of trastuzumab deruxtecan via real-world data of FDA adverse event reporting system database from 2019 to 2023

ABSTRACT Background Trastuzumab deruxtecan (T-DXd) is a novel generation antibody-drug conjugate. However, current evidence for the safety of T-DXd is mostly limited to clinical trials. It resulted in inability to timely update the adverse event (AE) for drug. Therefore, we mine and analyze AE signals caused by T-DXd based on the FAERS database, and to provide reference for the safe use of drug in clinical practice and reduce the occurrence of AE. Research design and methods This study analyzed real-world data on AEs associated with T-DXd reported in the FAERS database from December 2019 to December 2023, and the reporting odds ratio, proportional report ratio, bayesian confidence propagation neural network and multi-item gamma Poisson shrinker methods were combined to identify and evaluate reports, which can reduce the number of false positive and false negative signals with good sensitivity. Results A total of 4651 AE reports of T-DXd as the ‘primary suspect’ were collected. The median time to onset of T-DXd -related AEs was 25 days. The proportion of serious AEs was 69.40%. Hospitalization (25.67%) and death (24.19%) accounted for the highest occurrence frequency of clinical serious outcomes. The top three AE in both occurrence frequency and signal intensity were interstitial lung disease. We also found twelve kinds of potential novel danger signals not recorded in the drug lable, such as ascites, sepsis, pneumocystis jirovecii pneumonia, and gastroenteritis listeria, which need further attention and verification in clinical practice. Conclusion The results of this study indicate the importance of timely mining and updating information of adverse drug reactions. Drug evaluation should be performed well before using T-DXd, and during treatment, the indexes of pulmonary, bone marrow hematopoietic and cardiac function should be closely monitored, timely intervention measures should be taken to avoid related injury caused by AEs.

Symbolic computation of analytical solutions for nonlinear partial differential equations based on bilinear neural network method

Symbolic computation of analytical solutions for nonlinear partial differential equations based on bilinear neural network method

A High-Precision DOA Estimator for Low-Frequency Signals Using Ultra-Dense Small-Aperture Microphone Arrays

Abstract Accurate Direction of Arrival (DOA) estimation is critical for the effectiveness of Unattended Ground Sensor (UGS) systems, as it enhances sound localization, situational awareness, resource optimization, and facilitates integration with other sensor data for comprehensive monitoring. With the growing demand for lightweight and miniaturized sensors suitable for diverse environments, challenges arise in DOA estimation of low-frequency signals using dense small-aperture microphone arrays, especially under noisy conditions. Despite advancements in deep learning, both conventional and existing neural network methods struggle with this task. In this paper, we present the Multi-Resblock DOA Network (MRDNet), a novel neural network designed for precise DOA estimation of low-frequency sounds in noisy environments using small-aperture arrays. MRDNet was evaluated under simulations involving Brownian and Gaussian noise, representing wind and general background disturbances. The results demonstrate that MRDNet achieves superior accuracy, with a Mean Angular Error (MAE) of 2.797 degrees, significantly outperforming baseline methods by 24.93\%. Furthermore, we show that increasing the number of microphones within a constant array size using MRDNet effectively enhances DOA accuracy in the context of deep learning.

Novel Artificial Neural Network Aided Structural Topology Optimization

In this paper, novel artificial neural networks are adopted for the topology optimization of full structures at both coarse and fine scales. The novelty of the surrogate-based method is to use neural networks to optimize the relationship from boundary and mesh conditions to structure density distribution. The objective of this study is to explore the feasibility and effectiveness of deep learning techniques for structural topology optimization. The newly developed neural networks are used for optimizing various types of structures with different meshes, partition numbers, and parameters. The finite element computation takes more than 90% of the total operation time of the SIMP method, but it decreases to 40%. It is indicated that the computational cost for the whole structural design process is relatively low, while the accuracy is acceptable. The proposed artificial neural network method is used to perform topology optimization for some numerical examples such as the cantilever beam, the MBB beam, the L-shape beam, the column, and the rod-supported bridge. It is demonstrated that computational efficiency is considerably improved while the proposed neural network method is adopted.

Neural network algorithm for precipitation estimation from atms radiometer data

The paper presents a neural network method for precipitation estimation using microwave measurements from ATMS radiometer on board Suomi NPP and NOAA-20/21 satellites. The algorithms based on two fully-connected neural networks, the first one is used to detect precipitation clouds and the other one is used to quantify precipitation rate. When training the neural networks, the reference source of information was an array of measurements simulated using the fast radiation transfer model RTTOV in the bands of ATMS instrument and the corresponding precipitation rates were taken from ECMWF ERA5 reanalysis data. Validation of the obtained precipitation estimates was carried out using the results of the MIRS and GPROF algorithms for satellite radiometer ATMS, as well as ground-based radar observations from NIMROD. The results of the validation showed a high accuracy level consistent with many others works in this research field. The validation was carried out for land and water surface separately. The comparison with MIRS algorithm showed the correlation coefficient was more 0.9, and the RMSE error was approximately 0.78 mm/h for water and 0.84 mm/h for land surface. The same metrics for GPROF algorithm showed the correlation coefficient was ~0.8, and the RMSE error was approximately 1.27 mm/h and 0.9 for water and land surface, respectively. When compared with ground-based NIMROD radar data, the correlation and the RMSE were 0.47 and 1.37 mm/h, respectively. The results of the validation confirm the performance of the presented neural network method for precipitation estimation. In addition, further minor refinement of the presented algorithm will make it possible to apply it to measurements of other microwave satellite instruments, including Russian ones, such as MTVZA-GY, installed on Meteor-M satellites.

A Quality Prediction Method for Optimizing the Teaching Mode of History Courses Based on Multi-Category Data Fusion Analysis

In the process of predicting the optimization quality of teaching mode, a single convolutional neural network method is affected by multiple sources of data such as students’ behavioral data and teachers’ evaluations, which is prone to causing modal collapse and affecting the prediction quality. For this reason, we propose a method for predicting the optimization quality of history course teaching mode based on the fusion analysis of multi-class data. After collecting the data of optimizing the quality of teaching mode of historical courses, noise reduction is carried out by the noise reduction self-coder network on the data collection, and the combination of historical data and current data is realized by the dynamic slicing method of multi-feature matrix, after fusing the multi-category data, the data are inputted into the convolutional neural network, and an optimized multi-objective genetic algorithm optimizes the convolutional neural network, stabilizes the convolutional neural network model, and avoids the modal collapse caused by multi-source attributes and categories of the data. The optimized convolutional neural network is used to extract the deep features of the optimized quality of history course teaching mode, and the obtained features are input into the multiple linear regression model as independent variables to obtain the predicted scores of the optimized quality of history course teaching mode. Through experimental verification, the method can realize in-depth cleaning and denoising of the data of each feature of history curriculum teaching mode, and there is a large correlation between the features related to the optimization quality of the five history curriculum teaching modes collected by the method and the final score of the optimization quality of the history curriculum teaching modes, the method predicts that the optimization quality of the history curriculum teaching modes scores and the scores of the expert evaluation are basically consistent.

Triangular Mesh Surface Subdivision Based on Graph Neural Network

Mesh subdivision is a common mesh-processing algorithm used to improve model accuracy and surface smoothness. Its classical scheme adopts a fixed linear vertex update strategy and is implemented iteratively, which often results in excessive mesh smoothness. In recent years, a nonlinear subdivision method that uses neural network methods, called neural subdivision (NS), has been proposed. However, as a new scheme, its application scope and the effect of its algorithm need to be improved. To solve the above problems, a graph neural network method based on neural subdivision was used to realize mesh subdivision. Unlike fixed half-flap structures, the non-fixed mesh patches used in this paper naturally expressed the interior and boundary of a mesh and learned its spatial and topological features. The tensor voting strategy was used to replace the half-flap spatial transformation method of neural subdivision to ensure the translation, rotation, and scaling invariance of the algorithm. Dynamic graph convolution was introduced to learn the global features of the mesh in the way of stacking, so as to improve the subdivision effect of the network on the extreme input mesh. In addition, vertex neighborhood information was added to the training data to improve the robustness of the subdivision network. The experimental results show that the proposed algorithm achieved a good subdivision of both the general input mesh and extreme input mesh. In addition, it effectively subdivided mesh boundaries. In particular, using the general input mesh, the algorithm in this paper was compared to neural subdivision through quantitative experiments. The proposed method reduced the Hausdorff distance and the mean surface distance by 27.53% and 43.01%, respectively.

Is It Just About Scrolling? The Correlation of Passive Social Media Use with College Students’ Subjective Well-Being Based on Social Comparison Experiences and Orientation Assessed Using a Two-Stage Hybrid Structural Equation Modeling–Artificial Neural Network Method

Previous studies have found that passive social media use (PaSMU) tends to induce upward contrast, thereby affecting well-being. However, this perspective alone may overlook the mechanisms of other social comparison phenomena. This study analyzes the influence mechanism of PaSMU on subjective well-being (SWB) by categorizing social comparison into upward identification, upward contrast, downward identification, and downward contrast while incorporating social comparison orientation (SCO) as a moderating variable. This study surveyed college students who use RED (Xiaohongshu) and collected 352 valid questionnaires. A two-stage hybrid structural equation modeling (SEM)–artificial neural network (ANN) method was employed, utilizing path and mediation effect analysis to verify the moderating effect of SCO in the process of PaSMU affecting SWB. PaSMU is positively correlated with upward contrast and downward identification, both of which negatively affect SWB. Upward contrast and downward identification are associated with lower SWB, while downward comparison is positively correlated with SWB. High SCO strengthens the association between upward contrast and reduced SWB. Furthermore, upward contrast and downward identification were found to have comparable mediating effects between PaSMU and SWB. In contrast to previous studies, this research highlights that downward identification plays a comparably significant mediating role alongside upward contrast. Downward identification significantly mediates the relationship between PaSMU and SWB due to increased risk awareness, higher sensitivity to negative information among socially anxious students, emotional contagion from negative content, and anonymity that fosters an “imagined community”. Additionally, students with high SCO are more affected by idealized self-presentations and rely on upward contrasts for social feedback, lowering their SWB. This study reveals the complex correlation of PaSMU and SWB, providing new theoretical insights and practical strategies to encourage positive social media use among college students.

Fault diagnosis of inter‐turn short circuits in PMSM based on deep regulated neural network

AbstractPermanent Magnet Synchronous Machine (PMSM) is widely utilised in numerous industrial applications due to its precise control capabilities. However, these motors frequently encounter operational faults, potentially leading to severe safety and performance issues. Consequently, effective health monitoring techniques for early fault detection are essential to maintain optimal performance and extend the lifespan of these systems. This study presents a qualification‐based methodology for diagnosing faults in three‐phase PMSMs through vibration–current data fusion analysis. The stator faults, specifically inter‐turn short circuits (ITSC) induced via bypassing resistances, were investigated using experimental data from a custom‐built test rig. The collected current and vibration signals were transformed into statistical features. Various operating scenarios were diagnosed utilising a deep regulated neural network (RegNet), an improved convolutional neural network based on an enhanced residual architecture. The proposed approach was assessed through various metrics including training efficiency, precision, recall, f1‐score, and accuracy, and compared against several neural network methods. The findings reveal that the proposed RegNet model achieves perfect accuracy, attaining 100%. This research highlights the efficacy of data fusion analysis and deep learning in fault diagnosis, facilitating proactive maintenance strategies and improving the reliability of PMSMs in diverse industrial applications and renewable energy systems.

Sustainable Supplier Selection With Adaptive Network- Based Fuzzy Inference System (ANFİS)

Choosing the right sustainable supplier is a critical decision problem for businesses. Due to the increase in the number of competitors, the increase in evaluation criteria and the difficulty in defining evaluation criteria, the need for reliable methodologies that work with high accuracy increases as the complexity of the selection problem increases. This article discusses a new approach to the problem of sustainable supplier selection based on the Adaptive Network-Based Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) methods. In the study, firstly the sustainable supplier selection criteria have been reduced to four criteria by the ANFIS method. For sustainable supplier performance evaluation, ANFIS model and ANN model were developed. Multiple regression analysis was performed to compare the performance success of both models. The model with the highest success rate was determined as ANFIS model. At the end of the study, the most suitable sustainable supplier selection was made with ANFIS model.

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

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.

Real-Time On-Site Ground-Motion Prediction Using ResNet

Abstract On-site warnings can decrease the range of late alert zone during earthquakes. This study develops a deep learning model to predict whether the maximum peak ground acceleration at a station exceeds 25 Gal based on the waveform of the initial P-wave. A ResNet architecture model is developed to address the degradation problem in multilayered models, thereby enhancing performance. The model exhibits high performance for a test dataset, achieving F1-scores higher than 90% over various time windows, and outperforms the traditional progressive displacement value (Pdv) method and newer convolutional neural network (CNN) methods. Ground motion of two actual earthquakes are applied to test the model and validate its prediction capabilities, and the results indicate that the proposed model has higher accuracy and speed than Pdv and CNN methods. In addition, the model is integrated with the Earthworm earthquake early warning system for real-time waveform analysis and prediction, extending the model’s applicability beyond experimental stages to online systems. The proposed method is also tested using data from earthquakes that occurred in March 2023.

Implementasi Deep Learning Untuk Menentukan Harga Buah Sawit

This study aims to analyze the price of palm oil using Convolutional Neural Network (CNN) method in deep learning. CNN was chosen for its ability to process complex data and recognize patterns from diverse data. The stages of research include data analysis, data pre-processing, predictive model design for CNN method, CNN classification model prediction results, CNN method evaluation, and CNN method evaluation results. This study aims to produce a model that can predict the price of oil palm with high accuracy, based on data covering a variety of characteristics of farmers and the quality of oil palm crops. Prediction results were conducted using data from 50 oil palm farmers. From the prediction, as many as 23 data farmers get a price of IDR 2,300, 13 other farmers get a price of IDR 2,000, and the remaining 14 data farmers get a price of IDR 1,800. The results of this prediction are based on data from farmers and the quality of oil palm crops they grow and produce. By utilizing the CNN method, the model can capture various factors that affect the price of palm oil, including the quality of palm fruit and agricultural conditions. Evaluation of the CNN method showed very good results, with almost perfect accuracy. This method managed to predict palm oil prices very precisely, showing that CNN can be an effective tool in the analysis of palm oil prices. The results of this evaluation confirmed that the CNN method can be relied upon to provide accurate predictions, helping farmers and palm oil industry players in determining prices that are in accordance with the quality and condition of the crop.

Training improvement methods of ANN trajectory predictors in power systems

AbstractThis paper proposes training improvement methods of artificial neural networks (ANN) trajectory predictors. First, a dynamic power system time‐series trajectory is split into several different segments to simplify the original ANN training problem. Moreover, the time‐derivative of the trajectory is included to obtain an augmented loss function. Compared to previous studies which mainly focused on increasing the prediction accuracy, the aim of these novel techniques is to reduce the computational burden where the ANN output performance is still acceptable. The effectiveness of the developed methods is validated based on the WSCC three‐machine nine‐bus and IEEE 39‐bus system models. The mean absolute error (MAE) and trajectory prediction results are analysed, in which the numbers of neurons, hidden layers, and training epochs are constrained during the ANN training process. Rotor‐angle difference between generators and the system frequency are investigated as the dynamic trajectories of the power system models. The approaches are revealed to be effective when the ANN architecture and epochs are constrained. The MAE results can be reduced by up to 65% in the power system models depending on the ANN hyperparameters and training epochs. The ANN training results can better reflect the original trajectory as well.

Verification of uranium composite materials using active interrogation techniques for safeguards purposes

Abstract The uranium isotope characterization for nuclear safeguards and security purposes was conducted using both passive and active non-destructive techniques. The interrogated samples were natural uranium (NU), low-enriched uranium (LEU), and depleted uranium (DU) all in the form of uranium dioxide. The obtained spectra using both techniques were accumulated using a high-efficiency HPGe detector. Improving detector efficiency was developed using the neural network (NN) method based on the obtained intrinsic calibration data. The obtained enrichment results of passive measurements showed good agreement with certified samples to within 1.12% in the case of LEU, 1.03% in the case of NU, and 1.04% in the case of DU.
Active interrogation of the samples was done using the neutron spectrum of a 5 Ci 241Am-Be neutron source. Seven short-lived fission products (FPs) (101Tc, 97Nb, 105Ru, 92Y, 91Sr, 92Sr, and 88Kr) were chosen as indicators to provide a complete characterization of isotopic mass determination, uranium content, and enrichment. Isotopic masses of the samples were calculated using thermal and fast interrogations. A comparison of the obtained results of 238U mass was found to agree within 0.2% with certified value based on fast interrogation, and within 0.3% of 235U mass based on thermal interrogation.

Small aircraft detection in infrared aerial imagery based on deep neural network

Detection of aerial target is an important part of infrared image processing. Both neural network method and traditional method can be used in infrared object detection. Neural network method has many advantages such as high accuracy and good portability compared with traditional object detection method. Since the features extracted by neural network method can change over detection target, automatic feature extraction makes neural network based detection method more effective. In recent years deep learning method has been also found wide use for object detection in images. In this paper, an object detection model based on the deep learning network YOLO is constructed for solving the infrared aircraft detection problem. We construct the dataset used for training and testing with recognized features being iteratively learned. The task of infrared otject detection is sensitive to model size and detection speed. There is a requirement of using quantization method to reduce the storage space and the computation complexity. We propose a quantized model with appropriate accuracy for infrared object detection task. To solve the detection task for multiple extremely small aircrafts, model adjustment and quantization are used in proposed model and it gets a better performance. Experimental results on the constructed dataset show that the storage space for weight after quantization shrinks to a quarter, and there is no precision loss for extremely small aircrafts compared to the original model. The optimized YOLO-based deep learning model is effective to detect the small aircraft target in infrared aerial imagery.

Unified and evolved approach based on neural network and deep learning methods for intrusion detection

Currently, network security has become a major concern for all entities around the world. Attackers employ various methods to disrupt services, which requires new methods to stop them all in one way. Moreover, these intrusions can evolve and overcome security measures and devices, which pushes to use new evolving methods able to accompany the evolution of these threats, to block them. In our paper, we propose a new approach for intrusion detection, founded on neural network (NN) and deep learning (DL) methods. This approach is planned to not only identify threats, but also to develop a long-term memory of them, in order to detect new ones resembling these memorized attacks, and simultaneously, to provide a single way to stop all kinds of intrusions. To test our model, we have chosen the most recently employed methods in literature, NN and DL algorithms: feedforward neural network (FNN), convolutional neural network (CNN), and long short-term memory (LSTM), then we have applied them on network security layer-knowledge discovery in databases (NSL KDD) intrusions dataset. The results of experiments were impressive for all the algorithms, with maximum performances noted by LSTM, which affirms the efficacy of our proposed method for intrusion detection.