Neural Network Research Articles

Modified CNN Model for Network Intrusion Detection and Classification System Using Local Outlier Factor-based Recursive Feature Elimination

An intrusion detection system (IDS) is either a part of a software or hardware environment that monitors data and analyses it to identify any attacks made against a system or a network. Traditional IDS approaches make the system more complicated and less efficient, because the analytical properties process is difficult and time-consuming. This is because the procedure is complex. Therefore, this research work focuses on a network intrusion detection and classification (NIDCS) system using a modified convolutional neural network (MCNN) with recursive feature elimination (RFE). Initially, the dataset is balanced with the help of the local outlier factor (LOF), which finds anomalies and outliers by comparing the amount of deviation that a single data point has with the amount of deviation that its neighbors have. Then, a feature extraction selection approach named RFE is applied to eliminate the weakest features until the desired number of features is achieved. Finally, the optimal features are trained with the MCNN classifier, which classifies intrusions like probe, denial-of-service (DoS), remote-to-user (R2U), user-to-root (U2R), and identifies normal data. The proposed NIDCS system resulted in higher performance with 99.3% accuracy and a 3.02 false alarm rate (FAR) as equated to state-of-the-art NIDCS approaches such as deep neural networks (DNN), ResNet, and gravitational search algorithms (GSA).

Machine Learning based Wildfire Area Estimation Leveraging Weather Forecast Data

Wildfires are increasingly destructive natural disasters, annually consuming millions of acres of forests and vegetation globally. The complex interactions among fuels, topography, and meteorological factors, including temperature, precipitation, humidity, and wind, govern wildfire ignition and spread. This research presents a framework that integrates satellite remote sensing and numerical weather prediction model data to refine estimations of final wildfire sizes. A key strength of our approach is the use of comprehensive geospatial datasets from the IBM PAIRS platform, which provides a robust foundation for our predictions. We implement machine learning techniques through the AutoGluon automated machine learning toolkit to determine the optimal model for burned area prediction. AutoGluon automates the process of feature engineering, model selection, and hyperparameter tuning, evaluating a diverse range of algorithms, including neural networks, gradient boosting, and ensemble methods, to identify the most effective predictor for wildfire area estimation. The system features an intuitive interface developed in Gradio, which allows the incorporation of key input parameters, such as vegetation indices and weather variables, to customize wildfire projections. Interactive Plotly visualizations categorize the predicted fire severity levels across regions. This study demonstrates the value of synergizing Earth observations from spaceborne instruments and forecast data from numerical models to strengthen real-time wildfire monitoring and postfire impact assessment capabilities for improved disaster management. We optimize an ensemble model by comparing various algorithms to minimize the root mean squared error between the predicted and actual burned areas, achieving improved predictive performance over any individual model. The final metric reveals that our optimized WeightedEnsemble model achieved a root mean squared error (RMSE) of 1.564 km2 on the test data, indicating an average deviation of approximately 1.2 km2 in the predictions.

Metaheuristic-enhanced Deep Learning Model for Accurate Alzheimer's Disease Diagnosis from MRI Imaging

Alzheimer’s Disease (AD) is the neuro-degenerative dementia, where the precise and early recognition of AD is vital for timely treatment to reduce mortality rate. A new automated model is implemented in this work for early discovery of AD in the Magnetic Resonance Imaging (MRI) brain scans. Initially, the input brain scans are taken from the Alzheimer's disease Neuroimaging Initiative (ADNI) database. Further, the acquired raw brain scans are visually improved by employing the binary normalization technique. The denoised brain scans are fed to the pre-trained Convolutional Neural Network (CNN) named GoogleNet for feature extraction. Next, the extracted richer feature values are fed to the Long Short Term Memory (LSTM) network for classifying the brain scan as Normal Control (NC), Mild Cognitive Impairment (MCI) and AD. In this manuscript, a Honey Badger Optimization Algorithm (HBOA) technique is incorporated with the LSTM networks for hyper-parameters optimization, where this procedure helps in diminishing the LSTM network’s complexity and computational time. The experimental results conducted on the ADNI database underscore the HBOA-based LSTM network's effectiveness, showcasing a remarkable mean classification accuracy of 97.83% in multi-class classification. Moreover, the sensitivity of HBOA based LSTM for AD/NC is 96.73% which is high when compared to the existing methodologies such as SVM with radial basis kernel function and NCSINs. This performance surpasses that of other comparative models for AD detection, emphasizing the superior capabilities and potential of the proposed method in the early detection.

Machine learning model using immune indicators to predict outcomes in early liver cancer

BACKGROUND Patients with early-stage hepatocellular carcinoma (HCC) generally have good survival rates following surgical resection. However, a subset of these patients experience recurrence within five years post-surgery. AIM To develop predictive models utilizing machine learning (ML) methods to detect early-stage patients at a high risk of mortality. METHODS Eight hundred and eight patients with HCC at Beijing Ditan Hospital were randomly allocated to training and validation cohorts in a 2:1 ratio. Prognostic models were generated using random survival forests and artificial neural networks (ANNs). These ML models were compared with other classic HCC scoring systems. A decision-tree model was established to validate the contribution of immune-inflammatory indicators to the long-term outlook of patients with early-stage HCC. RESULTS Immune-inflammatory markers, albumin-bilirubin scores, alpha-fetoprotein, tumor size, and International Normalized Ratio were closely associated with the 5-year survival rates. Among various predictive models, the ANN model generated using these indicators through ML algorithms exhibited superior performance, with a 5-year area under the curve (AUC) of 0.85 (95%CI: 0.82-0.88). In the validation cohort, the 5-year AUC was 0.82 (95%CI: 0.74-0.85). According to the ANN model, patients were classified into high-risk and low-risk groups, with an overall survival hazard ratio of 7.98 (95%CI: 5.85-10.93, P < 0.0001) between the two cohorts. CONCLUSION A non-invasive, cost-effective ML-based model was developed to assist clinicians in identifying high-risk early-stage HCC patients with poor postoperative prognosis following surgical resection.

Bio-plausible reconfigurable spiking neuron for neuromorphic computing.

Biological neurons use diverse temporal expressions of spikes to achieve efficient communication and modulation of neural activities. Nonetheless, existing neuromorphic computing systems mainly use simplified neuron models with limited spiking behaviors due to high cost of emulating these biological spike patterns. Here, we propose a compact reconfigurable neuron design using the intrinsic dynamics of a NbO2-based spiking unit and excellent tunability in an electrochemical memory (ECRAM) to emulate the fast-slow dynamics in a bio-plausible neuron. The resistance of the ECRAM was effective in tuning the temporal dynamics of the membrane potential, contributing to flexible reconfiguration of various bio-plausible firing modes, such as phasic and burst spiking, and exhibiting adaptive spiking behaviors in changing environment. We used the bio-plausible neuron model to build spiking neural networks with bursting neurons and demonstrated improved classification accuracies over simplified models, showing great promises for use in more bio-plausible neuromorphic computing systems.

Classifying congenital brain anomalies using deep learning enabled Hunter-Squirrel Search Optimization

ABSTRACT Fetal brain anomalies are probably the most well-known inborn deformities that might be related to syndromic and chromosomal contortion. Early pre-birth location of cerebrum irregularities is fundamental for enhancing clinical management pathways and counselling for parents. To address this issue, Hunter Squirrel Search Optimization (HSSO)_LeNet enabled congenital brain anomaly classification is devised in this work. Commonly, an acquired input image is allowed to the preprocessing stage, which is accomplished by the Weiner filter. The preprocessed ultrasound image is allowed for the segmentation process, where image segmentation is done through 3D U-Net. Thereafter, the segmented image is allowed into the feature extraction process. Then, the brain anomaly classification is carried out using Deep Convolutional Neural Network (DCNN), where the hyperparameters are optimally tuned using the proposed HSSO. Consequently, the proposed HSSO algorithm is devised by incorporating the Hunter – Prey Optimizer (HPO) algorithm and the Squirrel Search Algorithm (SSA). Finally, the classification of congenital brain anomalies is carried out by LeNet, which is tuned by the proposed HSSO. The proposed technique has attained an accuracy of 93.79%, a True positive rate (TPR) of 94.67%, and a True Negative Rate (TNR) of 92.15%.

Optimized and improved DNN predictive model of fatigue life parameters in hybrid fiber reinforced self-compacting concrete

ABSTRACT Hybrid fiber-reinforced self-compacting concrete (HyFRSCC) represents a cutting-edge material in civil engineering, combining the benefits of self-compacting concrete with enhanced mechanical properties conferred by fiber reinforcement. As a versatile solution, HyFRSCC offers improved durability, sustainability, and structural performance compared to conventional concrete formulations. The novel techniques for improving the analysis and design of HyFRSCC structures with an emphasis on improving cost-effectiveness, sustainability, and durability in civil engineering applications. A novel approach that combines enhanced deep neural network (DNN)-based advanced predictive modeling with self-improved coati optimization (SI-COA) methodologies. By precisely predicting the Weibull distribution parameters which are essential for fatigue life analysis this method seeks to provide reliable characterization of HyFRSCC behavior under various stress conditions. Hybrid fibers, which frequently combine steel with synthetic or polymer fibers, improve concrete's resistance to cracking. The requirement for frequent maintenance or repairs is decreased by HFRSCC, which increases the longevity and durability of concrete structures. Hybrid fibers, which frequently combine steel with synthetic or polymer fibers, improve concrete's resistance to cracking. The concrete's ability to withstand freeze–thaw cycles is enhanced by fiber reinforcing. Fibers are added to concrete to improve its abrasion resistance and long-term performance under mechanical wear in applications where abrasion is an issue, such as industrial floors and pavements. When 60% of the training data is used, the Improved DNN model outperforms LSTM(47.00), DenseNet (44.70), MobileNet (49.49), ResNet (48.85), CNN (45.40), Bi-GRU (46.92), and DNN (42.52) with an MAE of 38.58.

Data augmentation and simultaneous prediction of optical properties of a porous core single mode fiber: CTGAN in the domain of optics

Abstract Photonic crystal fiber (PCF) architectures have garnered significant interest due to their versatile applications across various fields. However, the complexity of PCF designs and the computational demands of full vectorial finite element method (FV-FEM) simulations pose challenges in fully realizing their potential. While prior research has explored artificial neural networks (ANNs) to accelerate simulation predictions, proposed approaches were often limited by sample size and generalizability. In this manuscript we introduce conditional generative adversarial networks (CTGAN) to augment real datasets, facilitating more effective ANN training. We evaluate CTGAN’s performance by comparing the quality of augmented data with real data and assess the predictive accuracy of ANNs trained on both augmented and real datasets. Our results demonstrate enhanced predictive accuracy, with higher R 2 values for optical property predictions from structural parameters using the augmented dataset-trained ANN. Furthermore, the mean squared error (MSE) during ANN training decreased significantly (from 0.0051 to 0.0011), requiring fewer convergence epochs of only 88 compared to 114 for the real dataset. The proposed approach enables faster optical property predictions while reducing the required dataset generation simulations by up to 27.9 %.

Early stage cost prediction model for Indian building construction projects using artificial neural networks

Purpose This paper aims to enhance early-stage cost estimation in construction projects, a critical factor in project feasibility, funding, resource allocation and scheduling. Traditional cost estimation approaches suffer from limitations such as the absence of structured methodologies, assumptions of linear cost relationships, prolonged processes and expert judgment variations. To address these challenges, this study proposes a reliable cost prediction model based on artificial neural networks (ANNs) for building construction projects in India. Design/methodology/approach To develop cost prediction model, this study collected data from 377 building construction projects in India, encompassing 17 essential cost parameters. The methodology involves data preprocessing, constructing features and fine-tuning ANN hyperparameters meticulously to achieve optimal performance. Findings The research showcases effectiveness of cost prediction model, evident in significantly reduced mean square error values. ANN-based prediction model excels in handling nonlinear cost dependencies and diverse project complexities, making it a valuable tool for early-stage cost estimation. Research limitations/implications ANN-based cost prediction model is primarily designed for predicting costs associated with structural works of building projects. Practical implications The proposed solution offers stakeholders a robust data-driven decision-making tool during initial phases of construction projects. This can lead to more successful and economically viable outcomes. Originality/value This research examines the drawbacks of traditional cost estimation methods by presenting a data-driven approach leveraging machine learning. It significantly improves precision of early cost forecasts in construction projects while offering practical value to industry.

Detection of oscillation-like patterns in eclipsing binary light curves using neural network-based object detection algorithms

The primary aim of this research is to evaluate several convolutional neural network-based object detection algorithms for identifying oscillation-like patterns in light curves of eclipsing binaries. This involved creating a robust detection framework that can effectively process both synthetic light curves and real observational data. The study employs several state-of-the-art object detection algorithms, including Single Shot MultiBox Detector, Faster Region-based Convolutional Neural Network, You Only Look Once, and EfficientDet, as well as a custom non-pretrained model implemented from scratch. Synthetic light curve images and images derived from observational TESS light curves of known eclipsing binaries with a pulsating component were constructed with corresponding annotation files using custom scripts. The models were trained and validated on established datasets, which was followed by testing on unseen Kepler data to assess their generalisation performance. The statistical metrics were also calculated to review the quality of each model. The results indicate that the pre-trained models exhibit high accuracy and reliability in detecting the targeted patterns. The Faster Region-based Convolutional Neural Network and You Only Look Once in particular showed superior performance in terms of object detection evaluation metrics on the validation dataset, including a mean average precision value exceeding 99%. The Single Shot MultiBox Detector, on the other hand, is the fastest, although it shows a slightly lower performance, with a mean average precision of 97%. These findings highlight the potential of these models to significantly contribute to the automated determination of pulsating components in eclipsing binary systems and thus facilitate more efficient and comprehensive astrophysical investigations.

Sequence-dependent activity and compartmentalization of foreign DNA in a eukaryotic nucleus

In eukaryotes, DNA-associated protein complexes coevolve with genomic sequences to orchestrate chromatin folding. We investigate the relationship between DNA sequence and the spontaneous loading and activity of chromatin components in the absence of coevolution. Using bacterial genomes integrated into Saccharomyces cerevisiae , which diverged from yeast more than 2 billion years ago, we show that nucleosomes, cohesins, and associated transcriptional machinery can lead to the formation of two different chromatin archetypes, one transcribed and the other silent, independently of heterochromatin formation. These two archetypes also form on eukaryotic exogenous sequences, depend on sequence composition, and can be predicted using neural networks trained on the native genome. They do not mix in the nucleus, leading to a bipartite nuclear compartmentalization, reminiscent of the organization of vertebrate nuclei.

Simulation Research on the Optimization of Rural Tourism System Resilience Based on a Long Short-Term Memory Neural Network—Taking Well-Known Tourist Villages in Heilongjiang Province as Examples

Taking well-known tourist villages in Heilongjiang Province as the research object, we constructed a rural tourism system resilience assessment framework with the dimensions of “environment, institution, economy, society, and culture”. Using a geographical detector to analyze driving factors, an LSTM neural network model was constructed to predict the evolution trend of the rural tourism system resilience of these villages. The resulting insights included the following: ① The rural tourism system resilience of the well-known tourist villages in Heilongjiang Province is at a medium level, with a relatively good degree of development in the environmental dimension and the lowest degree in the economic dimension. ② The existence of financial support, water supply guarantee, domestic waste treatment, livestock manure treatment, and tourism development satisfaction are core driving factors for rural tourism system resilience; there is a non-linear or two-factor enhancement effect among these factors, and the interaction between domestic waste treatment and tourism development satisfaction has the strongest influence, while policy support particularly improves rural tourism system resilience and interacts most frequently with other driving factors. ③ Compared to the backpropagation (BP) neural network, the long short-term memory (LSTM) neural network has better stability and prediction accuracy.

Insilico, network pharmacology and neural network studies on Sida acuta Burm f.- based phytochemicals targeting NADH-ubiquinone oxidoreductase

The biochemical activities of phytochemicals obtained from S. acuta Burm f. have been described in several forms by various scientists globally. This work has revealed the antimalarial activities of S. acuta Burm f. using various methods such as density functional theory method, induced fit docking technique, quantitative structure activities relationship (QSAR) via neural network approach and pharmacokinetics studies using ADMETLab software. Compound F10 showed a stronger ability to interact in terms of HOMO energy and energy gap while Compound F12 exhibited a high tendency to accept electrons from nearby compounds. Furthermore, compound F13 demonstrated greater potential to inhibit NADH-Ubiquinone Oxidoreductase. We observed that the derivatives in F10, F12 and F13 enhanced the reactivity and inhibiting activities of the parent compound. The prediction of binding affinity for compounds F1 to F16 using neural network was found to be accurate and dependable. Also, ADMET study was executed and reported appropriately. Our research could pave the way for the creation of a library of effective phytochemicals from S. acuta Burm f -based drugs with potential anti-malaria activity.

Autoencoder artificial neural network for accelerated forward and inverse design of locally resonant acoustic metamaterials

The noise issues brought about by the development of the aviation and other industries have put forward an urgent demand for the design of low-frequency noise reduction structures. An autoencoder artificial neural network (ANN) is established in this paper to achieve accelerated low-cost forward and on demand design of locally resonant metamaterials simultaneously. Inspired by the framework of the autoencoder network, the proposed ANN is composed of an in series connected inverse prediction neural network and a forward prediction neural network module to avoid program errors by multisolution problems. A theoretical model is first set up in the paper to calculate the sound transmission loss (STL) of a locally resonant metamaterial plate and then validated by finite element simulation. The autoencoder ANN is subsequently trained using the dataset constructed based on the theoretical model. The accuracy of the well-trained ANN is then evaluated by making a comparison with the theoretical calculation and originally expected STL curves. The advantages of the proposed ANN over the theoretical model and numerical simulation are analyzed, and the results indicate that the proposed autoencoder ANN takes 2 and 6 orders of magnitude less time to complete the forward design than theoretical and numerical methods. The proposed ANN also demonstrates its ability in inverse design, which is hardly achieved using theoretical and numerical methods. The proposed ANN provides a new design method for accelerated forward and inverse design of noise reduction structures.

A Spatial‐Based Quantum Graph Convolutional Neural Network and Its Full‐Quantum Circuit Implementation

AbstractWith the rapid advancement of quantum computing, the exploration of quantum graph neural networks is gradually emerging. However, the absence of a circuit framework for quantum implementation and limited physical qubits hinder their realization on real quantum computers. To address these challenges, this paper proposes a spatial‐based quantum graph convolutional neural network and implements it on a superconducting quantum computer. Specifically, this model exclusively consists of quantum circuits, including quantum aggregation circuits in the quantum graph convolutional layer and quantum classification circuits in the quantum dense layer. To meet the requirements of Noisy Intermediate‐Scale Quantum computing, a first‐order extraction method to reduce circuit size is employed. Experimental results in node classification tasks demonstrate that this model achieves comparable or even superior performance compared to classical graph neural networks while utilizing fewer parameters. Therefore, this model can inspire further advancements in quantum graph neural networks and facilitate their implementation on physical quantum devices.

Forecasting Energy Consumption Trends with Machine Learning Models for Improved Accuracy and Resource Management in the USA

Accurate prediction of energy consumption patterns is vital for attaining sustainability and efficient economic planning. In the USA energy consumption trends have evolved substantially because of factors such as population growth, technological advancements, and shifts in consumption behaviors. The prime objective of this research project was to develop and evaluate machine learning algorithms with accuracy in predicting trends in America's consumption of energy. By employing complex methodologies such as neural networks, regression analysis, and ensemble approaches, this work aims to enhance the accuracy of forecasts in terms of demand for energy in residential, commercial, and industrial sectors. The U.S. consumption datasets covered a wide variety of information representing the consumption of electricity, use of fuel, and integration of renewable sources in residential, commercial, and transportation sectors. Most datasets used for analysis are taken from the U.S. Energy Information Administration (EIA) and the Department of Energy (DOE), offering in-depth statistics about production, consumption trends, and price trends. With our dataset consisting of continuous values for consumption of energy and many predictor factors, we compared a variety of machine algorithms, encompassing XG-Boost, Logistic Regression, and Random Forest. The implemented code compared three algorithms – Logistic Regression, Random Forest, and XG-Boost – in terms of performance via calculation and visualization of key evaluation metrics. It devised a function calculate-metrics to calculate accuracy, precision, recall, and F1-score for a prediction of each model over a test set. Retrospectively, comparing accuracy values, one can observe that Logistic Regression got a high score, with Random Forest following closely, and XG-Boost following a little behind them. Overall, through strategic plots, the comparative strengths and weaknesses of all three algorithms were seen, proving that Logistic Regression was the most reliable algorithm for predicting values in a dataset. Utility companies can benefit a lot through machine learning (ML)-based prediction in terms of distribution efficiency and overall operational efficiency. With ML algorithms, utility companies can utilize humungous volumes of consumption in the past to make future demand predictions with high accuracy.

Stability Analysis and Support Optimization of Tunnel Surrounding Rock with Weak Interlayer Based on Catastrophe Theory

A tunnel excavation model with weak interlayers is established to analyze the effect of different angles and thicknesses on the displacement of the surrounding rock. Based on catastrophe theory, a stability criterion for surrounding rock displacement is derived, providing a theoretical framework for evaluating tunnel stability through numerical simulations. Finally, to minimize support construction costs, an optimization model is established by integrating the Particle Swarm Optimization (PSO) algorithm and the Radial Basis Function (RBF) neural network. The model is then validated through engineering projects. The results show that the deformation and stability of the surrounding rock are affected by the interlayer angle and thickness. As the angle increases, the maximum deformation of the surrounding rock gradually transitions from horizontal displacement to vertical displacement, while increased thickness amplifies deformation and concentrates it at the interlayer. The stability of the surrounding rock exhibits catastrophic characteristics, with instability often occurring in the middle and rear sections of the tunnel. Calculation costs are reduced by 88% using the PSO-RBF optimization model, and construction costs decreased by 34.96% after optimizing the support parameters. This study provides theoretical support for the stability analysis and construction optimization of tunnels with weak interlayers.

Automating Prostate Cancer Grading: A Novel Deep Learning Framework for Automatic Prostate Cancer Grade Assessment using Classification and Segmentation.

Prostate Cancer (PCa) is the second most common cancer in men and affects more than a million people each year. Grading prostate cancer is based on the Gleason grading system, a subjective and labor-intensive method for evaluating prostate tissue samples. The variability in diagnostic approaches underscores the urgent need for more reliable methods. By integrating deep learning technologies and developing automated systems, diagnostic precision can be improved, and human error minimized. The present work introduces a three-stage framework-based innovative deep-learning system for assessing PCa severity using the PANDA challenge dataset. After a meticulous selection process, 2699 usable cases were narrowed down from the initial 5160 cases after extensive data cleaning. There are three stages in the proposed framework: classification of PCa grades using deep neural networks (DNNs), segmentation of PCa grades, and computation of International Society for Urological Pathology (ISUP) grades using machine learning classifiers. Four classes of patches were classified and segmented (benign, Gleason 3, Gleason 4, and Gleason 5). Patch sampling at different sizes (500 × 500 and 1000 × 1000 pixels) was used to optimize the classification and segmentation processes. The segmentation performance of the proposed network is enhanced by a Self-organized operational neural network (Self-ONN) based DeepLabV3 architecture. Based on these predictions, the distribution percentages of each cancer grade within the whole slide images (WSI) were calculated. These features were then concatenated into machine learning classifiers to predict the final ISUP PCa grade. EfficientNet_b0 achieved the highest F1-score of 83.83% for classification, while DeepLabV3 + architecture based on self-ONN and EfficientNet encoder achieved the highest Dice Similarity Coefficient (DSC) score of 84.9% for segmentation. Using the RandomForest (RF) classifier, the proposed framework achieved a quadratic weighted kappa (QWK) score of 0.9215. Deep learning frameworks are being developed to grade PCa automatically and have shown promising results. In addition, it provides a prospective approach to a prognostic tool that can produce clinically significant results efficiently and reliably. Further investigations are needed to evaluate the framework's adaptability and effectiveness across various clinical scenarios.

УПРАВЛЕНИЕ ТЕХНОЛОГИЧЕСКИМ ПРОЦЕССОМ КОПЧЕНИЯ РЫБНОЙ ПРОДУКЦИИ НА ОСНОВЕ ОЦЕНКИ ЕЕ КАЧЕСТВА

The purpose of the study is a control system for the technological process of smoking fish products based on instrumental assessment of its quality. The object of research is the technological process of smoking fish products. Theoretical and practical research was carried out on the basis of the department of Management of Technical Systems of the Federal State Budgetary Educational Institution of Higher Education Dalrybvtuz (Vladivostok, Primorsky Region). Organoleptic evaluation scales were built according to the recommendations of GOSTISO 11036-2017 “Organoleptic analysis. Methodology. Characteristics of the structure". Models were built using IDEF0 methodology tools. Organoleptic and instrumental methods for assessing the quality of smoked fish products, methods of structural and functional modeling of processes were used. Research was carried out in several stages: development of a methodology for measuring the color characteristics of smoked fish products, identification of ranges of values of the color characteristics of smoked fish products, development of a model for controlling the technological process of smoking fish products, development of a model for controlling the technological process of smoking fish products, development of a control system for the technological process of smoking fish products. The results of the study made it possible to develop patent № 217961 “Device for controlling the readiness of smoked fish products”, priority date – 12/16/2022. The developed method can serve as the basis for automation of both developed and existing smoking equipment. Its use minimizes the subjectivity of organoleptic methods for monitoring the readiness of smoked fish products, optimizes the duration of the process of processing fish raw materials with smoking means and thereby the energy consumption of the technological process. Further development of this research is possible by improving the data analysis stage at the stage of quality control of smoked fish products using a neural network, which, by analyzing color characteristics data, will offer the operator to vary the ranges of color characteristics values. The relevance of the proposals of a recurrent neural network can be realized using Markov chains.

Simultaneous approximation by neural network operators with applications to Voronovskaja formulas

AbstractIn this paper, we considered the problem of the simultaneous approximation of a function and its derivatives by means of the well‐known neural network (NN) operators activated by the sigmoidal function. Other than a uniform convergence theorem for the derivatives of NN operators, we also provide a quantitative estimate for the order of approximation based on the modulus of continuity of the approximated derivative. Furthermore, a qualitative and quantitative Voronovskaja‐type formula is established, which provides information about the high order of approximation that can be achieved by NN operators. To prove the above theorems, several auxiliary results involving sigmoidal functions have been established. At the end of the paper, noteworthy examples have been discussed in detail.