Stability Analysis of the Modified Levenberg-Marquardt Algorithm for the Artificial Neural Network Training

Modeling and optimization of docosahexaenoic acid production by Schizochytrium sp. based on kinetic modeling and genetic algorithm optimized artificial neural network

Docosahexaenoic acid (DHA), an essential ω-3 polyunsaturated fatty acid, is efficiently biosynthesized by Schizochytrium sp., yet its bioprocess optimization remains constrained by dynamic interdependencies between cultivation parameters and metabolic shifts. This study establishes a framework integrating kinetic modeling and machine learning to improve DHA production. Kinetic models based on Logistic and Luedeking-Piret equations were utilized to describe dynamic biomass, lipid and DHA production. An artificial neural network (ANN) trained on fermentation data predicted biomass and DHA yield, while genetic algorithm (GA) optimization elevated predictive accuracy (R2 = 0.988) and overcame local optimization. The ANN-GA model identified optimal three-stage control strategy, experimentally validating a 10.4 % increase in DHA yield (45.13 g/L) compared to optimal training data. By combining kinetic models and the ANN-GA model, this study provided a scalable framework for improving DHA production and reducing experimental costs.

Lasso Model-Based Optimization of CNC/CNF/rGO Nanocomposites

This study explores the use of citric acid and L-ascorbic acid as reducing agents in CNC/CNF/rGO nanocomposite fabrication, focusing on their effects on electrical conductivity and mechanical properties. Through comprehensive analysis, L-ascorbic acid showed superior reduction efficiency, producing rGO with enhanced electrical conductivity up to 2.5 S/m, while citric acid offered better CNC and CNF dispersion, leading to higher mechanical stability. The research employs an advanced optimization framework, integrating regression models and a neural network with 30 hidden layers, to provide insights into composition-property relationships and enable precise material tailoring. The neural network model, trained on various input variables, demonstrated excellent predictive performance, with R2 values exceeding 0.998. A LASSO model was also implemented to analyze variable impacts on material properties. The findings, supported by machine learning optimization, have significant implications for flexible electronics, smart packaging, and biomedical applications, paving the way for future research on scalability, long-term stability, and advanced modeling techniques for these sustainable, multifunctional materials.

Preparation and optimisation of solid lipid nanoparticles of rivaroxaban using artificial neural networks and response surface method

AIMS

This study aimed to improve rivaroxaban delivery by optimising solid lipid nanoparticles (SLN) for minimal mean diameter and maximal entrapment efficiency (EE), enhancing solubility, bioavailability, and the ability to cross the blood-brain barrier.

METHODS

A central composite design was employed to synthesise 32 SLN formulations. Response surface methodology (RSM) and artificial neural networks (ANN) models predicted mean diameter and EE based on five independent variables.

RESULTS

The optimised SLN formulation achieved a mean particle diameter of 159.8 ± 15.2 nm, with a Polydispersity index of 0.46, a zeta potential of -28.8 mV, and an EE of 74.3% ± 5.6%. The ANN model showed superior accuracy for both mean diameter and EE, outperforming the RSM model. Structural integrity and stability were confirmed by scanning electron microscopy (SEM), differential scanning calorimetry (DSC), and Fourier-transform infrared spectroscopy (FTIR).

CONCLUSION

The high accuracy of the ANN model highlights its potential in optimising pharmaceutical formulations and improving SLN-based drug delivery systems.

Statistical and machine learning models for location-specific crop yield prediction using weather indices

Crop yield prediction gains growing importance for all stakeholders in agriculture. Since the growth and development of crops are fully connected with many weather factors, it is inevitable to incorporate meteorological information into yield prediction mechanism. The changes in climate-yield relationship are more pronounced at a local level than across relatively large regions. Hence, district or sub-region-level modeling may be an appropriate approach. To obtain a location- and crop-specific model, different models with different functional forms have to be explored. This systematic review aims to discuss research papers related to statistical and machine-learning models commonly used to predict crop yield using weather factors. It was found that Artificial Neural Network (ANN) and Multiple Linear Regression were the most applied models. Support Vector Regression (SVR) model has a high success ratio as it performed well in most of the cases. The optimization options in ANN and SVR models allow us to tune models to specific patterns of association between weather conditions of a location and crop yield. ANN model can be trained using different activation functions with optimized learning rate and number of hidden layer neurons. Similarly, the SVR model can be trained with different kernel functions and various combinations of hyperparameters. Penalized regression models namely, LASSO and Elastic Net are better alternatives to simple linear regression. The nonlinear machine learning models namely, SVR and ANN were found to perform better in most of the cases which indicates there exists a nonlinear complex association between crop yield and weather factors.

Detection of moisture content of polyester fabric based on hyperspectral imaging and BP neural network

To validate the feasibility and improve the accuracy of water content detection in polyester fabrics using hyperspectral imaging, 150 sets of hyperspectral images of polyester fabrics with varying thicknesses and moisture contents were obtained, and the characteristics of the spectral curves and impact of moisture content were elucidated. In addition, the area and full width at half maximum of the characteristic peaks around 1363 and 1890 nm were determined as spectral characteristic variables. Furthermore, the models of polyester fabric moisture content detection were developed using backpropagation neural networks, and their accuracy was evaluated using correlation coefficient and mean squared error. It was observed that the change in the moisture content of polyester fabrics not only affected the reflectance of the overall spectral curve of polyester fabrics but also altered the position and overall shape of the characteristic peaks. As the moisture content increased, the proportion of pure water spectra in the mixed spectra of water-containing polyester fabrics also increased, leading to a change in the overall shape of the characteristic peaks of polyester fabrics. Because of the overlap between the near-infrared absorption bands of pure water and the polyester fabric around 1363 and 1890 nm, the area and full width at half maximum of the characteristic peaks were considered to be more representative than the reflection for modeling. The established backpropagation neural network-based moisture content quantitative detection model has shown extremely high detection accuracy, with the correlation coefficient for the test set being higher than 0.999 and the root mean square error being lower than 0.3 %, indicating that the detection error of moisture content was only about 0.3 wt%.

Predictive microbiology through the last century. From paper to Excel and towards AI

This chapter provides a historical perspective on predictive microbiology: from its inception till its current state, and including potential future developments. A look back to its origins in the 1920s underlies that scientists at the time had great ideas that could not be developed due to the lack of proper technologies. Indeed, predictive microbiology advancements mostly halted till the 1980s, when computing machines became broadly available, evidencing how these technologies were an enabler of predictive microbiology. Nowadays, predictive microbiology is a mature scientific field. There is a general consensus on experimental and computational methodologies, with software tools implementing these principles in a user-friendly manner. As a result, predictive microbiology is currently a useful tool for researchers, food industries and food safety legislators. On the other hand, this methodology has some important limitations that would be hard to solve without a reconsideration of some of its basic principles. In this sense, Artificial Intelligence and Data Science present great promise to advance predictive microbiology even further. Nevertheless, this would require the development of a novel conceptual framework that accommodates these novel technologies into predictive microbiology.

CNN-Optimized Electrospun TPE/PVDF Nanofiber Membranes for Enhanced Temperature and Pressure Sensing

Temperature and pressure sensors currently encounter challenges such as slow response times, large sizes, and insufficient sensitivity. To address these issues, we developed tetraphenylethylene (TPE)-doped polyvinylidene fluoride (PVDF) nanofiber membranes using electrospinning, with process parameters optimized through a convolutional neural network (CNN). We systematically analyzed the effects of PVDF concentration, spinning voltage, tip-to-collector distance, and flow rate on fiber morphology and diameter. The CNN model achieved high predictive accuracy, resulting in uniform and smooth nanofibers under optimal conditions. Incorporating TPE enhanced the hydrophobicity and mechanical properties of the nanofibers. Additionally, the fluorescent properties of the TPE-doped nanofibers remained stable under UV exposure and exhibited significant linear responses to temperature and pressure variations. The nanofibers demonstrated a temperature sensitivity of -0.976 gray value/°C and pressure sensitivity with an increase in fluorescence intensity from 537 a.u. to 649 a.u. under 600 g pressure. These findings highlight the potential of TPE-doped PVDF nanofiber membranes for advanced temperature and pressure sensing applications.

Machine vision-based autonomous road hazard avoidance system for self-driving vehicles

The resolution of traffic congestion and personal safety issues holds paramount importance for human's life. The ability of an autonomous driving system to navigate complex road conditions is crucial. Deep learning has greatly facilitated machine vision perception in autonomous driving. Aiming at the problem of small target detection in traditional YOLOv5s, this paper proposes an optimized target detection algorithm. The C3 module on the algorithm's backbone is upgraded to the CBAMC3 module, introducing a novel GELU activation function and EfficiCIoU loss function, which accelerate convergence on position loss lbox, confidence loss lobj, and classification loss lcls, enhance image learning capabilities and address the issue of inaccurate detection of small targets by improving the algorithm. Testing with a vehicle-mounted camera on a predefined route effectively identifies road vehicles and analyzes depth position information. The avoidance model, combined with Pure Pursuit and MPC control algorithms, exhibits more stable variations in vehicle speed, front-wheel steering angle, lateral acceleration, etc., compared to the non-optimized version. The robustness of the driving system's visual avoidance functionality is enhanced, further ameliorating congestion issues and ensuring personal safety.

Partial hard occluded target reconstruction of Fourier single pixel imaging guided through range slice

Fourier single pixel imaging utilizes pre-programmed patterns for laser spatial distribution modulation to reconstruct intensity image of the target through reconstruction algorithms. The approach features non-locality and high anti-interference performance. However, Poor image quality is induced when the target of interest is occluded in Fourier single pixel imaging. To address the problem, a deep learning-based image inpainting algorithm is employed within Fourier single pixel imaging to reconstruct partially obscured targets with high quality. It applies a distance-based segmentation method to segment obscured regions and the target of interest. Additionally, it utilizes an image inpainting network that combines multi-scale sparse convolution and transformer architecture, along with a reconstruction network that integrates Channel Attention Mechanism and Attention Gate modules to reconstruct complete and clear intensity images of the target of interest. The proposed method significantly expands the application scenarios and improves the imaging quality of Fourier single pixel imaging. Simulation and real-world experimental results demonstrate that the proposed method exhibits the high inpainting and reconstruction capacity in the conditions of hard occlusion and down-sampling.

Optimizing ultrashort pulse in fiber laser based on artificial intelligence algorithm

Ultrashort pulses, characterized by their short pulse duration, diverse spectral content, and high peak power, are widely used in fields including laser processing, optical storage, biomedical sciences, and laser imaging. The complex, highly-nonlinear process of ultrashort pulse evolution within fiber lasers is influenced by numerous aspects such as dispersion, loss, gain, and nonlinear effects. Traditionally, the split-step Fourier transforms method is employed for simulating ultrashort pulses in fiber lasers, which involves traversing multiple parameters within the fiber to attain the pulse's optimal state. The simulation is a significantly time-consuming process. Here, we use a neural network model to fit and predict the impact of multiple parameters on the pulse characteristics within fiber lasers, enabling parameter optimization through genetic algorithms to determine the optimal pulse duration, pulse energy, and peak power. Integrating artificial intelligence algorithms simplifies the acquisition of optimal pulse parameters and enhances our understanding of multiple parameters' impact on the pulse characteristics. The investigation of ultrashort pulse optimization based on artificial intelligence holds immense potential for laser design.

A contactless monitoring system for accurately predicting energy expenditure during treadmill walking based on an ensemble neural network

The monitoring of treadmill walking energy expenditure (EE) plays an important role in health evaluations and management, particularly in older individuals and those with chronic diseases. However, universal and highly accurate prediction methods for walking EE are still lacking. In this paper, we propose an ensemble neural network (ENN) model that predicts the treadmill walking EE of younger and older adults and stroke survivors with high precision based on easy-to-obtain features. Compared with previous studies, the proposed model reduced the estimation error by 13.95% and 66.20% for stroke survivors and younger adults, respectively. Furthermore, a contactless monitoring system was developed based on Kinect, mm-wave radar, and ENN algorithms, and the treadmill walking EE was monitored in real time. This ENN model and monitoring system can be combined with smart devices and treadmill, making them suitable for evaluating, monitoring, and tracking changes in health during exercise and in rehabilitation environments.

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

OBJECTIVE

A 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 METHODS

In 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.

RESULTS

The 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).

CONCLUSION

The neural network-based model for predicting OAR dose during IMRT for NPC is simple, reliable and worth further investigation and application.

Influence of the ANN Hyperparameters on the Forecast Accuracy of RAC's Compressive Strength

The artificial neural networks (ANNs)-based model has been used to predict the compressive strength of concrete, assisting in creating recycled aggregate concrete mixtures and reducing the environmental impact of the construction industry. Thus, the present study examines the effects of the training algorithm, topology, and activation function on the predictive accuracy of ANN when determining the compressive strength of recycled aggregate concrete. An experimental database of compressive strength with 721 samples was defined considering the literature. The database was used to train, validate, and test the ANN-based models. Altogether, 240 ANNs were trained, defined by combining three training algorithms, two activation functions, and topologies with a hidden layer containing 1-40 neurons. The ANN with a single hidden layer including 28 neurons, trained with the Levenberg-Marquardt algorithm and the hyperbolic tangent function, achieved the best level of accuracy, with a coefficient of determination equal to 0.909 and a mean absolute percentage error equal to 6.81%. Furthermore, the results show that it is crucial to avoid the use of overly complex models. Excessive neurons can lead to exceptional performance during training but poor predictive ability during testing.

Analysis and Prediction of Electrospun Nanofiber Diameter Based on Artificial Neural Network

Electrospinning technology enables the fabrication of electrospun nanofibers with exceptional properties, which are highly influenced by their diameter. This work focuses on the electrospinning of polyacrylonitrile (PAN) to obtain PAN nanofibers under different processing conditions. The morphology and size of the resulting PAN nanofibers were characterized using scanning electron microscopy (SEM), and the corresponding diameter data were measured using Nano Measure 1.2 software. The processing conditions and corresponding nanofiber diameter data were then inputted into an artificial neural network (ANN) to establish the relationship between the electrospinning process parameters (polymer concentration, applied voltage, collecting distance, and solution flow rate), and the diameter of PAN nanofibers. The results indicate that the polymer concentration has the greatest influence on the diameter of PAN nanofibers. The developed neural network prediction model provides guidance for the preparation of PAN nanofibers with specific dimensions.

Dose prediction of organs at risk in patients with cervical cancer receiving brachytherapy using needle insertion based on a neural network method

OBJECTIVE

A neural network method was employed to establish a dose prediction model for organs at risk (OAR) in patients with cervical cancer receiving brachytherapy using needle insertion.

METHODS

A total of 218 CT-based needle-insertion brachytherapy fraction plans for loco-regionally advanced cervical cancer treatment were analyzed in 59 patients. The sub-organ of OAR was automatically generated by self-written MATLAB, and the volume of the sub-organ was read. Correlations between D2cm³ of each OAR and volume of each sub-organ-as well as high-risk clinical target volume for bladder, rectum, and sigmoid colon-were analyzed. We then established a neural network predictive model of D2cm³ of OAR using the matrix laboratory neural net. Of these plans, 70% were selected as the training set, 15% as the validation set, and 15% as the test set. The regression R value and mean squared error were subsequently used to evaluate the predictive model.

RESULTS

The D2cm³/D90 of each OAR was related to volume of each respective sub-organ. The R values for bladder, rectum, and sigmoid colon in the training set for the predictive model were 0.80513, 0.93421, and 0.95978, respectively. The ∆D2cm³/D90 for bladder, rectum, and sigmoid colon in all sets was 0.052 ± 0.044, 0.040 ± 0.032, and 0.041 ± 0.037, respectively. The MSE for bladder, rectum, and sigmoid colon in the training set for the predictive model was 4.779 × 10^-3, 1.967 × 10^-3 and 1.574 × 10^-3, respectively.

CONCLUSION

The neural network method based on a dose-prediction model of OAR in brachytherapy using needle insertion was simple and reliable. In addition, it only addressed volumes of sub-organs to predict the dose of OAR, which we believe is worthy of further promotion and application.

Asynchronous dissipative stabilization for stochastic Markov-switching neural networks with completely- and incompletely-known transition rates

The asynchronous dissipative stabilization for stochastic Markov-switching neural networks (SMSNNs) is investigated. The aim is to design an output-feedback controller with inconsistent mode switching to ensure that the SMSNN is stochastically stable with extended dissipativity. Two situations, which involve completely- and incompletely-known transition rates (TRs), are taken into account. The situation that all TRs are exactly known is considered first. By applying a mode-dependent Lyapunov-Krasovskii functional, Dynkin's formula, and several matrix inequalities, a criterion for the desired performance of the closed-loop SMSNN is derived and a design method for determining the asynchronous controller is developed. Then, the study is generalized to the situation where some TRs are allowed to be uncertain or even fully unknown. An inequality is established for judging the upper bound of the product of the TRs with the Lyapunov matrix by making full use of accessible information on the incompletely-known TRs. Based on the inequality, performance analysis and control synthesis are presented without imposing the zero-sum hypothesis of the uncertainties in the TR matrix. Finally, an example with numerical calculation and simulation is provided to verify the validity of the stabilizing approaches.

Development of prediction models to estimate extubation time and midterm recovery time of ophthalmic patients undergoing general anesthesia: a cross-sectional study

BACKGROUND

To develop prediction models for extubation time and midterm recovery time estimation in ophthalmic patients who underwent general anesthesia.

METHODS

Totally 1824 ophthalmic patients who received general anesthesia at Joint Shantou International Eye Center were included. They were divided into a training dataset of 1276 samples, a validation dataset of 274 samples and a check dataset of 274 samples. Up to 85 to 87 related factors were collected for extubation time and midterm recovery time analysis, respectively, including patient factors, anesthetic factors, surgery factors and laboratory examination results. First, multiple linear regression was used for predictor selection. Second, different methods were used to develop predictive models for extubation time and midterm recovery time respectively. Finally, the models' generalization abilities were evaluated using a same check dataset with MSE, RMSE, MAE, MAPE, R-Squared and CCC.

RESULTS

The fuzzy neural network achieved the highest R-Squared of 0.956 for extubation time prediction and 0.885 for midterm recovery time, and the RMSE value was 6.637 and 9.285, respectively.

CONCLUSION

The fuzzy neural network developed in this study had good generalization performance in predicting both extubation time and midterm recovery time of ophthalmic patients undergoing general anesthesia.

TRIAL REGISTRATION

This study is prospectively registered in the Chinese Clinical Trial Registry, registration number: CHiCRT2000036416, registration date: August 23, 2020.

Vision measuring technology for the position degree of a hole group

The hole is one of the most important geometric elements in mechanical parts. The center distance of a hole group measurement method based on machine vision is proposed for solving the influence of perspective distortion and improving the applicability of vision systems. In the method, the plane equation of the measured plane is obtained by the line structured light vision technology, and the process is free from the constraints of the calibration plate. In order to eliminate the effect of projection distortion on the measurement accuracy, a local coordinate system is established on the plane of the measured hole group, the hole diameter, and the center distance of the hole group, which could be calculated by the local coordinates of the hole edge points. In the experiment, the flange is taken as the measured object, the distances between the holes on the flange are obtained by the method proposed in this paper, and the measurement results compared with the data are obtained by a coordinate measuring machine (CMM). The experimental results show that the average measurement error of center distance is 0.0739 mm, and the standard deviation is 0.0489 mm.

Utilization of artificial intelligence approach for prediction of DLP values for abdominal CT scans: A high accuracy estimation for risk assessment

Purpose

This study aimed to evaluate Artificial Neural Network (ANN) modeling to estimate the significant dose length product (DLP) value during the abdominal CT examinations for quality assurance in a retrospective, cross-sectional study.

Methods

The structure of the ANN model was designed considering various input parameters, namely patient weight, patient size, body mass index, mean CTDI volume, scanning length, kVp, mAs, exposure time per rotation, and pitch factor. The aforementioned examination details of 551 abdominal CT scans were used as retrospective data. Different types of learning algorithms such as Levenberg-Marquardt, Bayesian and Scaled-Conjugate Gradient were checked in terms of the accuracy of the training data.

Results

The R-value representing the correlation coefficient for the real system and system output is given as 0.925, 0.785, and 0.854 for the Levenberg-Marquardt, Bayesian, and Scaled-Conjugate Gradient algorithms, respectively. The findings showed that the Levenberg-Marquardt algorithm comprehensively detects DLP values for abdominal CT examinations. It can be a helpful approach to simplify CT quality assurance.

Conclusion

It can be concluded that outcomes of this novel artificial intelligence method can be used for high accuracy DLP estimations before the abdominal CT examinations, where the radiation-related risk factors are high or risk evaluation of multiple CT scans is needed for patients in terms of ALARA. Likewise, it can be concluded that artificial learning methods are powerful tools and can be used for different types of radiation-related risk assessments for quality assurance in diagnostic radiology.

A more cost-efficient Chinese Named Entity Recognition based on trigger and matching network

The lack of training data in new domain is a typical problem for named entity recognition (NER). Currently, researchers have introduced “entity trigger” to improve the cost-effectiveness of the model. However, it still required the annotator to attach additional trigger label, which increases the workload of the annotator. Moreover, this trigger applies only to English text and lacks research into other languages. Based on this problem, we have proposed a more cost-effective trigger tagging method and matching network. The approach not only automatic tagging entity triggers based on the characteristics of Chinese text, but also adds mogrifier LSTM to the matching network to reduce context-free representation of input tokens. Experiments on two public datasets show that our automatic trigger is effective. And it achieves better performances with automatic trigger than other state-of-the-art methods (The F1-scores increased by 1∼4).

A Novel Genetic Neural Network Algorithm with Link Switches and Its Application in University Professional Course Evaluation

This study exploits a novel enhanced genetic neural network algorithm with link switches (EGA-NNLS) to model the professional university course evaluating system. Various indices should be employed to evaluate the learning effect of a professional course comprehensively and objectively, and the traditional artificial evaluation methods cannot achieve this goal. The presented data-driven modeling method, EGA-NNLS, combines a neural network with link switches (NN-LS) with an enhanced genetic algorithm (EGA) and the Levenberg-Marquardt (LM) algorithm. It employs an optimized network structure combined with EGA and NN-LS to learn the relationships between the system's input and output from historical data and uses the network's gradient information via the LM algorithm. Compared with the traditional backpropagation neural network (BPNN), EGA-NNLS achieves a faster convergence speed and higher evaluation precision. In order to verify the efficiency of EGA-NNLS, it is applied to a collection of experimental data for modeling the professional university course evaluating system.

An Animation Model Generation Method Based on Gaussian Mutation Genetic Algorithm to Optimize Neural Network

With the rapid development of computer graphics, 3D animation has been applied to all fields of people's lives, especially in the industries of film and television works, games, and entertainment. The wide application of animation technology makes it difficult for general 3D animation effects to impress increasingly discerning audiences. Group animation, as a new focus, has received more and more attention and has become a hot issue in computer graphics. Traditional animation production mainly relies on manual drawing and key frame technologies. The limitations of these technologies make the production of group animation consume a lot of manpower, financial resources, and time, and cannot guarantee the intelligence of characters and the authenticity of group behavior. Therefore, in order to end the above issues, this paper proposes an animation model generation method based on Gaussian mutation genetic algorithm to optimize neural network, including obtaining animation scene data, according to the animation scene data, and extracting animation model elements. The elements are input into the model network, the target animation model is generated, and the target animation model is displayed. The method proposed in this paper improves the animation model generation method in the prior art to a certain extent. The proposed animation model is constructed only through fixed rules, and the composition rules of the model cannot be changed according to the historical data of the animation model construction and other factors. Technical issues that reduce the flexibility and accuracy of the animation model generation.

Estimation and Identification of Nonlinear Parameter of Motion Index Based on Least Squares Algorithm

Parameter identification is an important branch of automatic control. Due to its special function, it has been widely used in various fields, especially the modeling of complex systems or systems whose parameters are not easy to determine. With the development of control technology, the scale of the control object is getting larger and larger, which makes the calculation amount of the identification algorithm larger and larger. For the nonlinear system with complex structure, especially the nonlinear system containing the product of unknown parameters, the number of parameters of the over-parameterized identification method increases greatly, and the calculation amount of the identification algorithm also increases sharply. Therefore, a parameter estimation method with a small amount of calculation is explored. The results show that the proposed method can overcome the phenomenon of “data saturation”, thus improving the parameter identification results.

Full-convolution Siamese network algorithm under deep learning used in tracking of facial video image in newborns

This study was carried out with the aim of exploring the full-convolution Siamese network (SiamFC) in the application of neonatal facial video image tracking, achieving accurate recognition of neonatal pain and helping doctors evaluate neonatal emotions in an automatic manner. The current technology shows low accuracy on facial image recognition of newborns, so the SiamFC algorithm under the deep learning was optimized in this study. Besides, a newborn facial video image tracking model (FVIT model) was constructed based on the SiamFC algorithm in combination with the attention mechanism with face tracking algorithm, and the facial features of newborns were tracked and recognized. In addition, a newborn face database was constructed based on the adult face database to evaluate performance of the FVIT model. It was found that the accuracy of the improved algorithm is 0.889, higher by 0.036 in contrast to other models; the area under the curve (AUC) of success rate reaches 0.748, higher by 0.075 compared with other algorithms. What's more, the improved algorithm shows good performance in tracking the facial occlusion, facial expression changes, and scale conversion of newborns. Therefore, the improved algorithm shows higher accuracy and success rate and has good effect in capturing and tracking the facial images of newborns, thereby providing an experimental basis for facial recognition and pain assessment of newborns in the later stage.

AdjMix: simplifying and attending graph convolutional networks

Simple graph convolution (SGC) achieves competitive classification accuracy to graph convolutional networks (GCNs) in various tasks while being computationally more efficient and fitting fewer parameters. However, the width of SGC is narrow due to the over-smoothing of SGC with higher power, which limits the learning ability of graph representations. Here, we propose AdjMix, a simple and attentional graph convolutional model, that is scalable to wider structure and captures more nodes features information, by simultaneously mixing the adjacency matrices of different powers. We point out that the key factor of over-smoothing is the mismatched weights of adjacency matrices, and design AdjMix to address the over-smoothing of SGC and GCNs by adjusting the weights to matching values. Experiments on citation networks including Pubmed, Citeseer, and Cora show that our AdjMix improves over SGC by 2.4%, 2.2%, and 3.2%, respectively, while achieving same performance in terms of parameters and complexity, and obtains better performance in terms of classification accuracy, parameters, and complexity, compared to other baselines.

Diagnosis of early mild cognitive impairment using a multiobjective optimization algorithm based on T1-MRI data

Alzheimer's disease (AD) is the most prevalent form of dementia. The accurate diagnosis of AD, especially in the early phases is very important for timely intervention. It has been suggested that brain atrophy, as measured with structural magnetic resonance imaging (sMRI), can be an efficacy marker of neurodegeneration. While classification methods have been successful in diagnosis of AD, the performance of such methods have been very poor in diagnosis of those in early stages of mild cognitive impairment (EMCI). Therefore, in this study we investigated whether optimisation based on evolutionary algorithms (EA) can be an effective tool in diagnosis of EMCI as compared to cognitively normal participants (CNs). Structural MRI data for patients with EMCI (n = 54) and CN participants (n = 56) was extracted from Alzheimer's disease Neuroimaging Initiative (ADNI). Using three automatic brain segmentation methods, we extracted volumetric parameters as input to the optimisation algorithms. Our method achieved classification accuracy of greater than 93%. This accuracy level is higher than the previously suggested methods of classification of CN and EMCI using a single- or multiple modalities of imaging data. Our results show that with an effective optimisation method, a single modality of biomarkers can be enough to achieve a high classification accuracy.

Blending Colored and Depth CNN Pipelines in an Ensemble Learning Classification Approach for Warehouse Application Using Synthetic and Real Data

Electric companies face flow control and inventory obstacles such as reliability, outlays, and time-consuming tasks. Convolutional Neural Networks (CNNs) combined with computational vision approaches can process image classification in warehouse management applications to tackle this problem. This study uses synthetic and real images applied to CNNs to deal with classification of inventory items. The results are compared to seek the neural networks that better suit this application. The methodology consists of fine-tuning several CNNs on Red–Green–Blue (RBG) and Red–Green–Blue-Depth (RGB-D) synthetic and real datasets, using the best architecture of each domain in a blended ensemble approach. The proposed blended ensemble approach was not yet explored in such an application, using RGB and RGB-D data, from synthetic and real domains. The use of a synthetic dataset improved accuracy, precision, recall and f1-score in comparison with models trained only on the real domain. Moreover, the use of a blend of DenseNet and Resnet pipelines for colored and depth images proved to outperform accuracy, precision and f1-score performance indicators over single CNNs, achieving an accuracy measurement of 95.23%. The classification task is a real logistics engineering problem handled by computer vision and artificial intelligence, making full use of RGB and RGB-D images of synthetic and real domains, applied in an approach of blended CNN pipelines.

Predictive Maintenance Neural Control Algorithm for Defect Detection of the Power Plants Rotating Machines Using Augmented Reality Goggles

The concept of predictive and preventive maintenance and constant monitoring of the technical condition of industrial machinery is currently being greatly improved by the development of artificial intelligence and deep learning algorithms in particular. The advancement of such methods can vastly improve the overall effectiveness and efficiency of systems designed for wear analysis and detection of vibrations that can indicate changes in the physical structure of the industrial components such as bearings, motor shafts, and housing, as well as other parts involved in rotary movement. Recently this concept was also adapted to the field of renewable energy and the automotive industry. The core of the presented prototype is an innovative interface interconnected with augmented reality (AR). The proposed integration of AR goggles allowed for constructing a platform that could acquire data used in rotary components technical evaluation and that could enable direct interaction with the user. The presented platform allows for the utilization of artificial intelligence to analyze vibrations generated by the rotary drive system to determine the technical condition of a wind turbine model monitored by an image processing system that measures frequencies generated by the machine.

Estimation of Daily Reproduction Numbers during the COVID-19 Outbreak

(1) Background: The estimation of daily reproduction numbers throughout the contagiousness period is rarely considered, and only their sum R0 is calculated to quantify the contagiousness level of an infectious disease. (2) Methods: We provide the equation of the discrete dynamics of the epidemic’s growth and obtain an estimation of the daily reproduction numbers by using a deconvolution technique on a series of new COVID-19 cases. (3) Results: We provide both simulation results and estimations for several countries and waves of the COVID-19 outbreak. (4) Discussion: We discuss the role of noise on the stability of the epidemic’s dynamics. (5) Conclusions: We consider the possibility of improving the estimation of the distribution of daily reproduction numbers during the contagiousness period by taking into account the heterogeneity due to several host age classes.