Neural Network System Research Articles

A fast regional convolutional neural network system for the diagnosis of gastrointestinal subepithelial lesions on endoscopic ultrasound images.

A fast regional convolutional neural network system for the diagnosis of gastrointestinal subepithelial lesions on endoscopic ultrasound images.

Development of a Surrogate System of a Plant Dynamics Simulation Model and an Abnormal Situation Identification System for Nuclear Power Plants Using Deep Neural Networks

In the case of a new nuclear reactor, existing evaluation experience is limited; thus, accidents and troubles may occur as a result of such lack of experience. To deal with such situations, it is desirable to use a virtual nuclear plant to reproduce behaviors under various conditions and identify unknown anomalies from the behaviors. Then, when an abnormal situation occurs, one can quickly determine the cause of the abnormality to operate plant equipment and return the plant to a stable condition as quickly as possible. Two types of deep neural network (DNN) systems have been constructed to support the identification of unknown anomalies and the determination of their causes. One is a surrogate system that can estimate physical quantities of a nuclear power plant in a computational time of several orders less than a physical simulation model. The other is an abnormal situation identification system that can estimate the state of the disturbance causing an anomaly from physical quantities of a nuclear power plant. Both systems are trained and tested using data obtained from the analytical code for incore and plant dynamics (ACCORD), which reproduces the steady and dynamic behavior of the actual High Temperature Engineering Test Reactor (HTTR) under various scenarios. The DNN models are built by adjusting the main hyperparameters. Through these procedures, these systems are shown to be able to perform with a high degree of accuracy.

Reef fish identification through image records using artificial intelligence and convolutional neural network system

Coral reef environments show great diversity and abundance of species that represent important biological resources for obtaining food and medicines, in addition to acting as bioindicators of the quality of reef environments, serving as a parameter to diagnose environmental impacts and, simultaneously, to evaluate the availability and use of fishing stocks and biological resources in such environments. Monitoring such ecosystems arises from the urgent need to better understand the natural geographic and environmental variability of these systems, as well as their main drivers of change, to inform and promote more effective management of such environments. With the growing need to monitor communities of reef organisms, particularly fish, the creation of a digital analysis method would make a great contribution to the studies of reef fish. To fill this knowledge gap, our work evaluates the possibility of using an artificial intelligence mechanism using convolutional neural networks (CNN) to identify and count reef fish in image records, to optimize digital analysis of videos and/or photos in reef environments for fish ecology studies. The accuracy of the CNN identification method was effective, although it still has challenges to improve, such as: the need for a preliminary survey of reef fish species existing in the target area to be monitored, with the capture of good resolution images in different positions space for each species; production of an image bank with a large number of diverse images for each species (minimum of 200 images per species) and; To detect fish in videos, fast computers with a dedicated GPU card are needed, especially if the videos are high or ultra-resolution and have a large number of fps.

Investigation of Hydrodynamics Characteristics of Fluidized Bed with back propagation Artificial Neural Network (BPANN)

Correlations have also been developed with system parameters by using dimensional analysis and an artificial neural network approach. The paper describes an investigation for the thermal design of a fluidized bed cooler and prediction of heat transfer rate among the media categories. It is devoted to the thermal design of such equipment and their application in the industrial fields. In the present work, an extensive ANN by using back propagation (BP) has been carried out to correlate the expansion ratio, fluctuation ratio in gas-solid fluidized bed. Back propagation network is the most well-known and widely used among the current types of neural network system, several applications of ANN for modeling of nonlinear process systems and subsequent control were reported. In back-propagation, different ANN structures (I×H×O) with varying number of neurons in the hidden layer were used as a tool for training input and output data for prediction value of hydrodynamic characteristics of the bed system. It was noted that ring models are the best ones in reducing bed expansion ratio and fluctuation ratio around (25%) and (23.22%).

Insights into the prediction of the liquid density of refrigerant systems by artificial intelligent approaches

This study presents a novel model for accurately estimating the densities of 48 refrigerant systems, categorized into five groups: Hydrofluoroethers (HFEs), Hydrochlorofluorocarbons (HCFCs), Perfluoroalkylalkanes (PFAAs), Hydrofluorocarbons (HFCs), and Perfluoroalkanes (PFAs). Input variables, including pressure, temperature, molecular weight, and structural groups, were systematically considered. The study explores the efficacy of both the multilayer perceptron artificial neural network (MLP-ANN) and adaptive neuro-fuzzy inference system (ANFIS) methodologies in constructing a precise model. Utilizing a comprehensive dataset of 3825 liquid density measurements and outlier analysis, the models achieved R2 and MSE values of 0.975 & 0.5575 and 0.967 & 0.7337 for MLP-ANN and ANFIS, respectively, highlighting their remarkable predictive performance. In conclusion, the ANFIS model is proposed as an effective tool for estimating refrigerant system densities, particularly advantageous in scenarios where experimental measurements are resource-intensive or sophisticated analysis is required.

Pattern dynamics analysis of a time-space discrete FitzHugh-Nagumo (FHN) model based on coupled map lattices

This paper investigates the dynamics of a discrete FitzHugh-Nagumo (FHN) model with self-diffusion on two-dimensional coupled map lattices. The primary objective is to analyze the complex dynamics of neuronal systems in a discrete setting. Through the application of central manifold and normal form analysis, it has been demonstrated that the system is capable of undergoing Neimark-Sacker and flip bifurcations even in the absence of diffusion. Additionally, when influenced by diffusion, the system can manifest pure Turing instability, Neimark-Sacker-Turing instability, and Flip-Turing instability. In the numerical simulation section, the path from bifurcation to chaos is explored by calculating the maximum Lyapunov exponent and drawing the bifurcation diagram. The interconversion between various Turing instabilities is simulated by varying the values of the time step and the self-diffusion coefficient. This study contributes to a deeper understanding of the complexity of neural network systems.

Predicting the Compressive Strength of Environmentally Friendly Concrete Using Multiple Machine Learning Algorithms

Machine learning (ML) algorithms have been widely used in big data prediction and analysis in terms of their excellent data regression ability. However, the prediction accuracy of different ML algorithms varies between different regression problems and data sets. In order to construct a prediction model with optimal accuracy for fly ash concrete (FAC), ML algorithms such as genetic programming (GP), support vector regression (SVR), random forest (RF), extremely gradient boost (XGBoost), backpropagation artificial neural network (BP-ANN) and adaptive network-based fuzzy inference system (ANFIS) were selected as regression and prediction algorithms in this study; the particle swarm optimization (PSO) algorithm was also used to optimize the structure and hyperparameters of each algorithm. The statistical results show that the performance of the assembled algorithms is better than that of an NN-based algorithm. In addition, PSO can effectively improve the prediction accuracy of the ML algorithms. The comprehensive performance of each model is analyzed using a Taylor diagram, and the PSO-XGBoost model has the best comprehensive performance, with R2 and MSE equal to 0.9072 and 11.4546, respectively.

Machine learning for power transformer SFRA based fault detection

This paper presents machine learning methods for health assessment of power transformer based on sweep frequency response analysis. The paper presents an overview of monitoring and diagnostics based on statistical Sweep Frequency Response Analysis (SFRA) based indicators that are used to evaluate the state of the power transformer. Experimental data obtained from power transformers with internal short-circuit faults is used as a database for applying machine learning. Machine learning is implemented to achieve more precise asset management and condition-based maintenance. Unsupervised machine learning was applied through the k-means cluster method for classifying and dividing the examined power transformer state into groups with similar state and probability of failure. Artificial neural network (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) as part of supervised machine learning are created in order to detect fault severity in tested power transformers of different lifetime. The presented machine learning methods can be used to improve health assessment of power transformers.

Finite-Time Passivity and Synchronization for a Class of Fuzzy Inertial Complex-Valued Neural Networks with Time-Varying Delays

This article investigates finite-time passivity for fuzzy inertial complex-valued neural networks (FICVNNs) with time-varying delays. First, by using the existing passivity theory, several related definitions of finite-time passivity are illustrated. Consequently, by adopting a reduced-order method and dividing complex-valued parameters into real and imaginary parts, the proposed FICVNNs are turned into first-order real-valued neural network systems. Moreover, appropriate controllers and the Lyapunov functional method are established to obtain the finite-time passivity of FICVNNs with time delays. Furthermore, some essential conditions are established to ensure finite-time synchronization for finite-time passive FICVNNs. In the end, corresponding simulations certify the feasibility of the proposed theoretical outcomes.

On the design and the digital implementation of observer‐based controllers for tracking of nonlinear time‐delay systems

AbstractIn this article, the tracking control problem for a class of nonlinear time‐delay systems is investigated. In particular, a new methodology for the design and the digital implementation of observer‐based tracking controllers is provided for a class of control‐affine nonlinear systems with state delays. First, a procedure for the design of continuous‐time observer‐based tracking controllers ensuring the global asymptotic stability of the corresponding closed‐loop tracking error system is provided for the considered class of systems. Then, sufficient conditions are provided for the existence of a suitably fast sampling and of an accurate quantization of the input/output channels such that the digital implementation of the proposed continuous‐time observer‐based tracking controller ensures the semi‐global practical stability property of the related sampled‐data quantized closed‐loop tracking error system, with arbitrarily small final target ball of the origin. Moreover, it is shown that, in the special case of delay‐free nonlinear systems, the sufficient conditions provided for the digital implementation of the proposed continuous‐time observer‐based tracking controller can be strongly relaxed. In the theory here developed, time‐varying sampling periods and nonuniform quantization of the input/output channels are taken into account. The proposed results are validated through examples concerning a class of neural networks systems and a class of time‐delay systems including, as a special case, a delay‐free actuated inverted pendulum.

Machine learning, conventional and statistical physics modeling of 2,4-Dichlorophenoxyacetic acid (2,4-D) herbicide removal using biochar prepared from Vateria indica fruit biomass

The adsorption properties of 2,4-Dichlorophenoxyacetic acid (2,4-D) onto biochar, obtained through HCl-assisted hydrothermal carbonization process of Vateria indica fruits (VI-BC), were extensively studied using traditional and statistical physics approaches. The traditional adsorption investigations encompassed kinetics, equilibrium, and thermodynamics studies. Subsequently, the Hill statistical physics model was employed to interpret the mechanism. Also, artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) machine learning tools were successfully employed to model the adsorption data wherein both models had high prediction potential (R2 > 0.99). The outcomes demonstrated that the produced VI-BC exhibited remarkable adsorptive traits, having a considerable specific surface area (111.54 m2/g), pore size (5.89 nm), a variety of functional groups, and appropriate attributes for efficiently adsorbing 2,4-D. For 10 mg/L 2,4-D, at pH 2.0 and with 0.3 g/L dose, an impressive 91.67% adsorption efficiency was achieved within a 120-min. Pseudo-second-order model aptly depicted the kinetic behavior of 2,4-D adsorption, while the Freundlich model provided a more accurate representation of the isotherms. 2,4-D maximum adsorption capacity stood at 131.39 mg/g at 303 K. The Hill statistical physics model elucidated that the adsorption primarily occurred via physisorption mechanisms, involving electrostatic attractions, π-π conjugation, and pore filling. This conclusion was further substantiated by post-adsorption characterization of the VI-BC. Thermodynamic analysis indicated that the interactions between VI-BC and 2,4-D were favorable, spontaneous, and exothermic. The calculated low energy of adsorption (1.255 kJ/mol) and ΔH° value (−20.49 kJ/mol) further supported physisorption as the dominant mechanism. In summary, this study underscores the significant potential of the newly developed biochar as a promising alternative material for efficiently removing the 2,4-D herbicide from polluted environments.

Morphology-based deep learning enables accurate detection of senescence in mesenchymal stem cell cultures.

Cell senescence is a sign of aging and plays a significant role in the pathogenesis of age-related disorders. For cell therapy, senescence may compromise the quality and efficacy of cells, posing potential safety risks. Mesenchymal stem cells (MSCs) are currently undergoing extensive research for cell therapy, thus necessitating the development of effective methods to evaluate senescence. Senescent MSCs exhibit distinctive morphology that can be used for detection. However, morphological assessment during MSC production is often subjective and uncertain. New tools are required for the reliable evaluation of senescent single cells on a large scale in live imaging of MSCs. We have developed a successful morphology-based Cascade region-based convolution neural network (Cascade R-CNN) system for detecting senescent MSCs, which can automatically locate single cells of different sizes and shapes in multicellular images and assess their senescence state. Additionally, we tested the applicability of the Cascade R-CNN system for MSC senescence and examined the correlation between morphological changes with other senescence indicators. This deep learning has been applied for the first time to detect senescent MSCs, showing promising performance in both chronic and acute MSC senescence. The system can be a labor-saving and cost-effective option for screening MSC culture conditions and anti-aging drugs, as well as providing a powerful tool for non-invasive and real-time morphological image analysis integrated into cell production.

НЕЙРОСЕТЕВАЯ СИСТЕМА РАСПОЗНАВАНИЯ ДОРОЖНЫХ ЗНАКОВ

This paper describes the stages and results of developing a neural network system for recognizing road signs. Current problems of image analysis and recognition find their effective solution in neural network methods, while artificial neural networks of convolutional architecture are leading in this subject area. One of the current and practically significant tasks in this area is the recognition of traffic signs. To solve it, it is proposed to use a convolutional neural network and a neural network system implemented on its basis. Describes the use of convolutional neural networks for image classification. To develop a neural network model, the initial stage of preparing a data set - a database of road signs - is described. The publicly available GTSRB set of traffic signs was used to train the neural network. The set contains 43 types of signs, the total number of images is about 39,000. To build a neural network model, 30,000 images of road signs were randomly selected from the original data. To validate the model - 9000 images. The Python programming language was used to develop a convolutional neural network system. In addition, the following libraries were used: OpenCV, NumPy, Scikit-learn, Scikit-image, Matplotlib, Imutils and TensorFlow. Anaconda was used as the development environment. The developed program includes three main files. The modelnet.py file describes the architecture of the neural network. The train.py file contains data processing and training of the neural network to classify traffic signs. The predict.py file presents an implementation of a ready-made convolutional neural network model for classifying road signs. The Adam optimization algorithm was used to directly train the neural network model. The neural network was trained using the fit method from the Keras library. As a result of training, the neural network achieved an accuracy of 94.7% on the training data set and an accuracy of 94.7% on the validation set. Through training, the neural network has learned to recognize and classify different signs effectively, which allows it to be used for automatic sign recognition in real images.

BOUALLEGUE EQUATION SYSTEM AND ITS APPLICATIONS IN CHAOS THEORY

This paper presents a new contribution in chaos theory using Bouallegue Equation System, this system is built of three blocks, These blocks are connected in cascade, the first block is a system of chaotic attractor, the second block is a system of fractal processes and the third block is a system of neural networks. In this work, we have generated a novel class of hyper chaotic attractors by coupling with neuron containing multi dendrites, each dendrite has its activation function. The activation function contains four parameters form, position of dendrite, parameter of equilibrium and polarity. Each dendrite can take four behaviors by configuration the parameter n, its structure contains four variables (n,p,q,χ), so we call this capacity or this potential of activation function, a Variable Structure Model of Neuron (VSMN, for short). The impact of activation function in chaotic attractor generates a new classes of hyper chaos attractors with fluctuation, cutting, folding, amplifying in attractors. Simulations have demonstrated the validity and feasibility of the proposed method. Two results of applications using Bouallegue Equation System in fractal theory are also given to show the similarities in medicine illustrating the fractal chromosome and the fractal lung, and in biology displaying a fractal flower.

Relaxed Stability Criteria for Delayed Memristor-Based Neural Network Systems via a Novel Matrix-Separation Legendre Inequality.

This article studies the issue of stability in memristor-based neural network (MNN) systems with time-varying delays. First, a novel matrix-separation Legendre inequality is proposed to achieve a tight hierarchical bound on augmented-type integral terms. To derive implementable inequality conditions, several delay-dependent matrices are introduced to eliminate the reciprocal terms associated with time-varying delay. Furthermore, a new Lyapunov-Krasovskii (L-K) functional is proposed by incorporating augmented-type double integrals and delay-product terms. A series of free-weighting matrices are introduced into the L-K functional, leveraging the zero-sum equations and the S-procedure pertaining to both the delay and its derivative. Based on the proposed matrix-separation Legendre inequality and L-K functional, the derived stability conditions exhibit reduced conservatism, as validated by three numerical cases and simulation results.

A Decentralized and Self-Adaptive Intrusion Detection Approach Using Continuous Learning and Blockchain Technology

The landscape of cyber threats is constantly in flux, which can cause traditional intrusion detection systems (IDS) to fall behind the rapid evolution of attacks. This can result in delayed detection and devastating consequences. The proposed system leverages continuous learning and self-adaptive neural networks to transcend the limitations of traditional IDS. It takes a proactive approach by continuously analyzing intrusion logs, using a Long Short-Term Memory (LSTM) core to identify emerging patterns and refine its understanding of threats in real time. This eliminates the static limitations of traditional models and encourages continuous improvement. The system's neural network is distributed across each block of the blockchain, creating a decentralized 'brain' that develops defenses against advanced adversaries. Secure enclaves are located within Trusted Execution Environments (TEEs), ensuring tamper-proof operation and reliable threat detection. The system's superior performance is demonstrated through rigorous evaluations of established datasets, such as NSL-KDD. The model demonstrates a high level of accuracy of 0.9994 with a minimal false positive rate of 0.06, indicating its ability to differentiate between legitimate network activity and malicious intrusions. Embracing continuous learning and a decentralized architecture creates a dynamic and resilient system that proactively adapts to the ever-changing threat landscape. This approach has several advantages over traditional solutions, including enhanced precision, increased security, and real-time adaptability.

Advancing Video Search Capabilities: Integrating Feedforward Neural Networks for Efficient Fragment-Based Retrieval

In the context of rapidly increasing volumes of video data, the problem of their efficient search and analysis becomes more acute. This research aims to develop and test an innovative system to improve the speed and accuracy of video search, utilizing the capabilities of Deep Convolutional Neural Networks (DCNN) and Feedforward Neural Networks (FFNN). Within the methodology developed for this study, video data are processed through several sequential stages: from feature extraction to key frame identification and the formation of an abstract vector representation. Deep Convolutional Neural Networks are central to the system for image analysis and Feedforward Neural Networks for optimizing the search process. The main results of the study include an increase in video search efficiency by reducing data processing time and increasing the accuracy of identifying relevant fragments. The originality of the work lies in the integration of two types of neural networks for structured analysis of video data, which is a new step in the development of video search technologies. The practical significance of the research is expressed in the possibility of applying the developed system in various areas where fast and accurate video search is needed: from the media industry to security systems. The scope of further research includes adapting the system to specific types of video content and expanding the capabilities of artificial intelligence for a deeper understanding of video data.

A novel rice plant leaf diseases detection using deep spectral generative adversarial neural network

The farming industry widely requires automatic detection and analysis of rice diseases to avoid wasting financial and other resources, reduce yield loss, improve processing efficiency, and obtain healthy crop yields. The proposed Deep Spectral Generative Adversarial Neural Network (DSGAN2) method is used for detecting rice plant leaf disease. Initially, fed into the input of healthy and non-healthy leaves from the collected dataset. Then, apply an Improved Threshold Neural Network (ITNN) method to enhance the image quality. Next, it uses a Segmentation using a Segment Multiscale Neural Slicing (SMNS) algorithm to identify the support-intensive color saturation based on the enhanced image. After that, the Spectral Scaled Absolute Feature Selection (S2AFS) method is applied to select optimal features and the closest weight from segmented rice plant leaves. Social Spider Optimization will select the feature using the Closest Weight (S2O-FCW) algorithm to analyze the feature weight values. Finally, the proposed Soft-Max Logistic Activation Function with Deep Spectral Generative Adversarial Neural Network (DSGAN2) algorithm detects rice plant disease based on selected features. With an accuracy of 97 %, the model helps farmers identify and identify Rice Plant diseases. The proposed system Deep Spectral Generative Adversarial Neural Network (DSGAN2) produces a decreasing false rate compared to the existing system of ACPSOSVM-Dual Channels Convolutional Neural Network (APS-DCCNN) is 55.2 %, Alex Net is 50.4 %, and Convolutional Neural Network (CNN) is 49.5 %.

Aquaculture monitoring system using multi-layer perceptron neural network and adaptive neuro fuzzy inference system

<span>The water quality is the most important parameter for aquatic species health and growth. The condition is very critical and is essential to monitor continuously. Poor water quality will affect health, growth and ability of the animal to survive. These also affected their harvesting yields based on the amount and size of the animal. The main water parameters such dissolved oxygen (DO), pH, temperature, salinity and turbidity are monitored and control for good water quality. The data were acquired by the developed instrument and send wirelessly through GPRS/GSM module to cloud-based database. The data were retrieved and the water quality is predicted using fuzzy logic and multi-layer perceptron. MATLAB software was used for the model which is developed based on Mamdani fuzzy interface system. The membership functions of fuzzy were generated, as well as the simulation and analysis of the water quality system. Results show that the performance of fuzzy method can improve system performance in monitoring the water quality. This system also provides alert signals to farmers based on specific limit value for the water quality parameters. This will help the breeders to make certain adjustment to ensure suitable water quality for the aquaculture system.</span>

Deep Neural Network and Radiomics-based Magnetic Resonance Imaging System for Predicting Microvascular Invasion in Hepatocellular Carcinoma

Deep Neural Network and Radiomics-based Magnetic Resonance Imaging System for Predicting Microvascular Invasion in Hepatocellular Carcinoma