Results Of Experiments Research Articles

Saturated Sliding Mode Control Scheme for a New Wearable Back-Support Exoskeleton

Low back pain has been torturing people around the world as a common and chronic disease, the main inducement of which is lumbar disc herniation (LDH). To alleviate patients’ pain noninvasively, this paper proposes a new lumbar spinal rehabilitation exoskeleton. This exoskeleton is composed of two bands connected by four motor-driven piston pushrods, where the range of band can be adjusted to adapt to people with different waistlines. The four pushrods provide support forces to hold the upper body to relieve the burden of lumbar. The joints between pushrods and band are universal pairs and spherical pairs. The whole structure can be regarded as a parallel robot, thus the supporting and waist tracking performance can be determined by motion control effect of pushrods. Considering the motor torque of pushrods is limited, in this paper, a saturated sliding mode control scheme is proposed. Meanwhile, an extended state observer (ESO) is employed to estimate the external disturbance and internal uncertainties, then the estimates of perturbations will be compensated by the feedback scheme. Furthermore, to enhance the disturbance-tracking performance of the ESO an integrated sliding mode observer is proposed to compensate the tracking errors of ESO. The stabilities of controller and observers are proved by Lyapunov stability theory. Finally, a simulation and two experiments are conducted to verify the performance of the proposed controller and the new exoskeleton. The simulation results show that the novel/new controller can drive the new exoskeleton to move along with the desirable trajectory to support the upper body so that alleviating the burden of waist. The results of myoelectricity experiments also show favorable effectiveness of exoskeleton on supporting upper body in different postures. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —this article was inspired by the concerns of low back pain that many people are suffering. Most existing rehabilitation robots tend to focus on supporting lower body and arms, while back-support rehabilitation robots are barely seen. Although we can see a few upper-body supporting robots in sporadic papers, these studies have not proposed a good control scheme yet. Besides, it is difficult to find a rehabilitation robot with simple structure as well as conforming to ergonomics. To resolve the shortcomings of the structure of existing robots and control models, we proposed a completely new rehabilitation robot with novel mechanical structure. Meanwhile, a new saturated sliding mode and novel integrated observer are employed in the position-force control system to enhance the stability and controllability of the whole robot control system. Then, some elaborately designed experiments are conducted to test the proposed robot and the control system. Numerical results demonstrate that the proposed method can significantly increase the trajectory tracking performance and alleviate low back pain.

Decentralized Scheduling for Concurrent Tasks in Mobile Edge Computing via Deep Reinforcement Learning

Mobile Edge Computing (MEC) is a promising solution to enhance the computing capability of resource-limited networks. A fundamental problem in MEC is efficiently offloading tasks from user devices to edge servers. However, there still exists a gap to deploy in real-world environments: 1) traditional centralized approaches needs complete information of edge network, ignoring the communication costs generated by synchronization, 2) previous works do not consider concurrent computation on edge servers, which may cause dynamic changes in the environment, and 3) the scheduling algorithm should deliver individualized decisions for different users independently and with high efficiency To solve this mismatch, we studied a multi-user task offloading problem where user devices make offloading decisions independently. We consider the concurrent execution of tasks and formulate a non-divisible and delay-aware task offloading problem to jointly minimize the dropped task ratio and long-term latency. We propose a decentralized task scheduling algorithm based on DRL that makes offloading decisions without knowing the information of other user devices. We employ Double-DQN, Dueling-DQN, Prioritized Replay Memory, and Recurrent Neural Network (RNN) techniques to improve the algorithm's performance. The results of simulation experiments show that our method can significantly reduce the long-term latency and dropped task ratio compared to the baseline algorithms.

Disturbance Observer-Based Fixed-Time Tracking Control for a Tilt Trirotor Unmanned Aerial Vehicle

This paper proposed a new fixed-time trajectory tracking control strategy based on the disturbance observer (DOB) for a tilt trirotor unmanned aerial vehicle (TTUAV) which is subject to unknown external disturbances. To facilitate the TTUAV's trajectory tracking controller design, the dynamic model of the TTUAV is divided into two subsystems, the position subsystem (outer-loop) and the attitude subsystem (inner-loop). For the dynamics of the TTUAV's outer-loop, the unknown external disturbances consist of the unknown wind disturbances, the unknown aerodynamic frictions forces, the lateral force generated by the rear servo and the states coupling between the outer-loop and the inner-loop. Then, a nonlinear DOB is designed to estimate the unknown external disturbances, and a nonsingular terminal sliding-mode controller (NTSMC) is further developed to drive the TTUAV's position tracking errors to zero in a fixed time. For the dynamics of the TTUAV's inner-loop, the unknown external disturbances are mainly introduced by the aerodynamic frictions torques. Then, a modified sliding mode controller is designed to reject the unknown external disturbances, and to ensure that the attitude tracking errors converge to zero exponentially. In addition, a composite stability analysis based on Lyapunov stability theory is provided to guarantee the fixed-time convergence of the closed-loop dynamics of the TTUAV. Finally, the results of the real-time experiments are presented to verify the effectiveness of the proposed control strategy.

Functional Connectivity Prediction With Deep Learning for Graph Transformation.

Inferring resting-state functional connectivity (FC) from anatomical brain wiring, known as structural connectivity (SC), is of enormous significance in neuroscience for understanding biological neuronal networks and treating mental diseases. Both SC and FC are networks where the nodes are brain regions, and in SC, the edges are the physical fiber nerves among the nodes, while in FC, the edges are the nodes' coactivation relations. Despite the importance of SC and FC, until very recently, the rapidly growing research body on this topic has generally focused on either linear models or computational models that rely heavily on heuristics and simple assumptions regarding the mapping between FC and SC. However, the relationship between FC and SC is actually highly nonlinear and complex and contains considerable randomness; additional factors, such as the subject's age and health, can also significantly impact the SC-FC relationship and hence cannot be ignored. To address these challenges, here, we develop a novel SC-to-FC generative adversarial network (SF-GAN) framework for mapping SC to FC, along with additional metafeatures based on a newly proposed graph neural network-based generative model that is capable of learning the stochasticity. Specifically, a new graph-based conditional generative adversarial nets model is proposed, where edge convolution layers are leveraged to encode the graph patterns in the SC in the form of a graph representation. New edge deconvolution layers are then utilized to decode the representation back to FC. Additional metafeatures of subjects' profile information are integrated into the graph representation with newly designed sparse-regularized layers that can automatically select features that impact FC. Finally, we have also proposed new post hoc explainer of our SF-GAN, which can identify which subgraphs in SC strongly influence which subgraphs in FC by a new multilevel edge-correlation-guided graph clustering problem. The results of experiments conducted to test the new model confirm that it significantly outperforms existing state-of-the-art methods, with additional interpretability for identifying important metafeatures and subgraphs.

Self-Reporting Microsensors Inspired by Noctiluca Scintillans for Small-Defect Positioning and Electrical-Stress Visualization in Polymers.

Small defects induce concentrated electrical stress in dielectric polymers, leading to premature failure of materials. Existing sensing methods fail to effectively visualize these defects owing to the invisible-energy state of the electric field. Thus, it is necessary to establish a nondestructive method for the real-time detection of small defects in dielectric polymers. In this study, a self-reporting microsensor (SRM) inspired by Noctiluca scintillans is designed to endow materials with the ability of self-detection for defects and electrical stress. The SRM leverages the energy of a nearby electric field to emit measurable fluorescence, enabling defect localization and diagnosis as well as electrical-stress visualization. A controllable dielectric microsphere is constructed to achieve an adjustable electroluminescence threshold for the SRM, thereby increasing its detection accuracy while decreasing the electroluminescence threshold. The potential degradation in the polymer performance owing to SRM implantation is addressed by assembling long molecular chains on the SRM surface to spontaneously generate an interpenetrating network. Results of finite element analyses and experiments demonstrate that the SRM can effectively realize nondestructive visualization and positioning of small defects and concentrated electrical stress in polymers, positioning it as a promising sensing method for monitoring the electric field and charge distribution in materials.

Introducing active sites and regulating electron distribution to design Mo and P co-doped Ni for efficient alkaline hydrogen evolution reaction

NiMoP electrocatalysts for hydrogen evolution reaction (HER) are received widespread concern, but the mechanism of Mo/P atoms for improving the HER activity still need to be revealed in details. Herein, we synthesize the Mo/P co-doped Ni (NMP) electrocatalysts, and the alkaline HER performances are tested. The NMP exhibit the excellent activity, durability and stability of material characters. Based on the results of experiments and density functional theory calculations, it is found that the introduced P atoms can regulate the electron redistribution, which can result in enhancing the reaction kinetic, increasing the active sites and charge exchange capacity. Besides, it leads the doped Mo to become the active site in both Volmer and Heyrovsky processes, and Ni bonded to Mo and Ni atoms exhibit the increasing activity. The obtained results display a deeper understanding for the doped metal and non-metallic atoms in enhancing HER activity of Ni, which is of great significant for designing the highly efficient and inexpensive electrocatalysts for producing hydrogen.

Inducing Clusters Deep Kernel Gaussian Process for Longitudinal Data

We consider the problem of predictive modeling from irregularly and sparsely sampled longitudinal data with unknown, complex correlation structures and abrupt discontinuities. To address these challenges, we introduce a novel inducing clusters longitudinal deep kernel Gaussian Process (ICDKGP). ICDKGP approximates the data generating process by a zero-mean GP with a longitudinal deep kernel that models the unknown complex correlation structure in the data and a deterministic non-zero mean function to model the abrupt discontinuities. To improve the scalability and interpretability of ICDKGP, we introduce inducing clusters corresponding to centers of clusters in the training data. We formulate the training of ICDKGP as a constrained optimization problem and derive its evidence lower bound. We introduce a novel relaxation of the resulting problem which under rather mild assumptions yields a solution with error bounded relative to the original problem. We describe the results of extensive experiments demonstrating that ICDKGP substantially outperforms the state-of-the-art longitudinal methods on data with both smoothly and non-smoothly varying outcomes.

An integrated model of supply chain resilience considering supply and demand uncertainties

AbstractThe complexity of the global supply chain has increased dramatically over the past few decades as a result of uncertainty caused by various factors. This paper studies the optimal strategy for supply chain resilience models considering supply disruptions and demand fluctuations. We present two‐stage stochastic programming models based on different scenarios, including a risk‐neutral model that considers the expected total cost, a risk‐averse model that considers the conditional value‐at‐risk measure, and a responsiveness model that considers the service level. We also propose multiobjective mathematical programming that considers all three models simultaneously and suggests the solution approach. Finally, we present the results of computational experiments and demonstrate how to cope with uncertainty through flexibility and redundancy. We offer a set of nondominated solutions from the multiobjective model and derive managerial insights, which suggest a decision‐making strategy between the disruption risk, expected total cost, and service level.

Research of the process of biological wastewater treatment under conditions of uneven load of the treatment system

The main purpose of the article is to develop a multifactorial model for rapid assessment of the efficiency of biological wastewater treatment reactors. A mathematical model of the process of biological wastewater treatment has been developed based on: changes in the concentration of organic contaminants in the bioreactor over time, taking into account the uneven flow of wastewater to the treatment plant, the process of substrate entering the bioreactor (different amounts may enter at different times). The software implementation of the proposed algorithm for solving the corresponding model problem in Python is carried out. The results of computer experiments on the study of the efficiency of wastewater treatment in biological treatment reactors for different operating conditions of facilities are presented. In particular, such processes were considered with taking into account the unevenness of the load, because the maximum cleaning loads are in the morning and in the evening. The task was solved to simulate a real situation and show how cleaning takes place at the maximum load at a certain time of the day. The results obtained will be useful for calculations in the design of biological treatment facilities or in the reconstruction of existing bioreactors for their prospective operation under new operating conditions.

Self‐Distancing Regulates the Effect of Incidental Anger (vs. Fear) on Affective Decision‐Making Under Uncertainty

ABSTRACTEmotions integral to a task are often adaptive, particularly in situations where outcomes and probabilities are not known. However, decisions are also influenced by emotions that arise from situations unrelated to the task. This is especially the case with negative emotions like fear and anger, which also tend to be accompanied by ruminative thinking that might divert decision‐makers' attention from the task at hand. In two preregistered experiments, we show how self‐distancing regulates the influence of incidental anger (vs. fear) on decision‐making under uncertainty. Participants recalled and reflected on a fear‐related or anger‐related event from either a self‐immersed or self‐distanced perspective. Next, they completed a task that is commonly used to measure affective decision‐making under uncertainty, the Iowa Gambling Task. The results in both experiments indicated that self‐immersed angry (vs. fearful) decision‐makers were significantly slower to avoid the risky, disadvantageous decks. These findings demonstrate how the ways in which we process negative emotional events shape their carryover effects in decision‐making under uncertainty and point to self‐distancing as a potential tool to control incidental emotional influences.

Preexposure to running attenuates rats' running-based flavour avoidance: Testing associative blocking with a cover cues or context change.

Voluntary running in activity wheels by rats leads to a Pavlovian conditioned aversion to the flavour consumed immediately before the running, causing the rats to avoid that flavour. This learning process, known as running-based flavour avoidance learning (FAL), is weakened when the rats have had repeated exposure to the wheels before. According to the associative account, the association between the background context and running established during the preexposure phase blocks the conditioning of the target flavour because the running is highly predictable by the background context from the outset of the FAL phase. Experiments 1 and 2 examined this account by introducing another flavour as a cue signalling wheel access during the preexposure phase. In the framework of the associative account, the introduction of this cue should impede the formation of the context-running association during the preexposure phase, thereby hindering the contextual blocking of aversive conditioning for the target flavour in the FAL phase. This would result in unweakened FAL. Although the results of Experiment 1 align with this prediction, in Experiment 2, when highly distinct flavours were used as the target and second cues, the preexposure effect was not eliminated. This contradicts the predictions of the associative account, indicating that Experiment 1 may have been influenced by stimulus generalisation. In Experiment 3, changing background contexts between the preexposure and FAL phases had no impact on the preexposure effect, contrary to the predictions of the associative account. In general, the associative account was not supported.

PDTE: Pyramidal deep Taylor expansion for optical flow estimation

Optical flow estimation is an important hot research in computer vision. Although existing methods had got a considerable progress in improving their performance, they still have drawbacks, such as heavily computational burden, inaccurate pixel-level offset estimation, and poor interpretability. To address these issues, this letter proposes a pyramidal deep Taylor expansion (PDTE) framework, including: First, we seriously interpreted the relationship between optical flow computation and Taylor expansion. Then, the proposed PDTE is constructed by employing global motion aggregation (GMA) to calculate each derivative part which contributes to the final estimated optical flow. Quantitative and qualitative results on the Sintel and KITTI datasets validate that the proposed PDTE scheme is effective and outperforms the state-of-the-art optical flow estimation methods. The results of extensive experiments in the ablation study demonstrate that PDTE performs well on shape preservation and the accuracy improvement of optical flow estimation, even pixel-level offset calculation.

Efficient Homolytic Cleavage of H2O2 on Hydroxyl‐Enriched Spinel CuFe2O4 with Dual Lewis Acid Sites

AbstractTraditional H2O2 cleavage mediated by macroscopic electron transfer (MET) not only has low utilization of H2O2, but also sacrifices the stability of catalysts. We present a non‐redox hydroxyl‐enriched spinel (CuFe2O4) catalyst with dual Lewis acid sites to realize the homolytic cleavage of H2O2. The results of systematic experiments, in situ characterizations, and theoretical calculations confirm that tetrahedral Cu sites with optimal Lewis acidity and strong electron delocalization can synergistically elongate the O−O bonds (1.47 Å → 1.87 Å) in collaboration with adjacent bridging hydroxyl (another Lewis acid site). As a result, the free energy of H2O2 homolytic cleavage is decreased (1.28 eV → 0.98 eV). H2O2 can be efficiently split into ⋅OH induced by hydroxyl‐enriched CuFe2O4 without MET, which greatly improves the catalyst stability and the H2O2 utilization (65.2 %, nearly 2 times than traditional catalysts). The system assembled with hydroxyl‐enriched CuFe2O4 and H2O2 affords exceptional performance for organic pollutant elimination. The scale‐up experiment using a continuous flow reactor realizes long‐term stability (up to 600 mL), confirming the tremendous potential of hydroxyl‐enriched CuFe2O4 for practical applications.

Magnetic Supramolecular Spherical Arrays: Direct Formation of Micellar Cubic Mesophase by Lanthanide Metallomesogens with 7-Coordination Geometry.

Here, an unprecedented phenomenon in which 7-coordinate lanthanide metallomesogens, which align via hydrogen bonds mediated by coordinated H2 O molecules, form micellar cubic mesophases at room temperature, creating body-centered cubic (BCC)-type supramolecular spherical arrays, is reported. The results of experiments and molecular dynamics simulations reveal that spherical assemblies of three complexes surrounded by an amorphous alkyl domain spontaneously align in an energetically stable orientation to form the BCC structure. This phenomenon differs greatly from the conventional self-assembling behavior of 6-coordinated metallomesogens, which form columnar assemblies due to strong intermolecular interactions. Since the magnetic and luminescent properties of different lanthanides vary, adding arbitrary functions to spherical arrays is possible by selecting suitable lanthanides to be used. The method developed in this study using 7-coordinate lanthanide metallomesogens as building blocks is expected to lead to the rational development of micellar cubic mesophases.

Molecular basis of distinct responses to drought between rice and wheat genotypes

The water-intensive cultivation of rice in flooded conditions poses a great challenge to global water resources, particularly in the face of climate change. Wheat requires significantly less water than rice with fairly high yields. A main research question is whether there are any drought adaptive mechanisms in rice similar to those existing in wheat. We report a study, where precise moisture stress was imposed at the vegetative stage using a high-throughput drought simulator phenomics facility. Variability in stress response was captured through untargeted metabolome analysis. In the first experiment, using two rice cultivars with contrasting drought responses (IR64 and Apo) and a wheat cultivar, Weebill, we found distinct variations in regulation of metabolite levels under well-watered or water-limited conditions, both between and within species. Several biosynthetic pathways, including those of polyamines, flavonoids, flavones, and specific metabolites like sphinganine-1-phosphate and anthranilate were enhanced under water-limited conditions. These metabolites were more abundant in wheat than in rice. Among the rice cultivars, Apo had higher abundance of metabolites than IR64. Regulatory patterns governing these metabolic pathways provided basis for stress responses of the tolerant rice cultivar and wheat. A second experiment was conducted to screen rice germplasm lines for acquired tolerance traits contrasting for drought responsiveness. Eight genotypes (four sensitive and four tolerant ones) were selected to validate the candidate metabolites from the first experiment. In the third experiment, an RT-PCR analysis of the key regulatory genes governing the metabolic pathways was conducted, using plant materials from both experiments. The results of this third experiment provided a strong validation for the relevance of these metabolic pathways under stress in rice. We identified promising rice genotypes that exhibited fold change levels of stress resistance metabolites comparable to those of the wheat cultivar Weebill. This study helped identify key candidate metabolites that can be either developed as markers or as elicitors, and genotypes that can be used as trait donor lines for rice improvement.

The grammaticalization and grammatical meaning of a ‘do’ support verb in the northern Italian Camuno dialect

Abstract This paper provides a case study of the grammaticalization of ‘do’ as a support verb in the northern Italian dialect of Camuno. It demonstrates the relative compatibility of fa ‘do’ with different semantic types of supported verbs through the results of an elicitation experiment. A comparison of results from adjacent dialects reveals a series of progressively more grammaticalized forms with three stages: ‘pragmatic’ ‘do’-support, largely restricted to agentive-subject, activity verbs, where question type depends mainly on the pragmatics; ‘agentive’ ‘do’-support, where use with that core group of verbs is largely categorial, and there is more use with theme-subject, change-of-state verbs and a few experiencer-subject, stative verbs; and ‘generalized’ ‘do’-support where optionality remains only with the most resistant stative verbs. Along with loss of lexical content in ‘do’, comes a reduction in salience of the special question meaning. Furthermore, a question with lexical fa ‘do’ (in an optional FS dialect) implies physical evidence exists to determine the truth of a proposition. In a question with bleached fa ‘do’ (in an obligatory FS dialect) the connection to reality has been lost leaving only the implication, and fa ‘do’ has then become an epistemic modal.

Multi-modal deep learning methods for classification of chest diseases using different medical imaging and cough sounds.

Chest disease refers to a wide range of conditions affecting the lungs, such as COVID-19, lung cancer (LC), consolidation lung (COL), and many more. When diagnosing chest disorders medical professionals may be thrown off by the overlapping symptoms (such as fever, cough, sore throat, etc.). Additionally, researchers and medical professionals make use of chest X-rays (CXR), cough sounds, and computed tomography (CT) scans to diagnose chest disorders. The present study aims to classify the nine different conditions of chest disorders, including COVID-19, LC, COL, atelectasis (ATE), tuberculosis (TB), pneumothorax (PNEUTH), edema (EDE), pneumonia (PNEU). Thus, we suggested four novel convolutional neural network (CNN) models that train distinct image-level representations for nine different chest disease classifications by extracting features from images. Furthermore, the proposed CNN employed several new approaches such as a max-pooling layer, batch normalization layers (BANL), dropout, rank-based average pooling (RBAP), and multiple-way data generation (MWDG). The scalogram method is utilized to transform the sounds of coughing into a visual representation. Before beginning to train the model that has been developed, the SMOTE approach is used to calibrate the CXR and CT scans as well as the cough sound images (CSI) of nine different chest disorders. The CXR, CT scan, and CSI used for training and evaluating the proposed model come from 24 publicly available benchmark chest illness datasets. The classification performance of the proposed model is compared with that of seven baseline models, namely Vgg-19, ResNet-101, ResNet-50, DenseNet-121, EfficientNetB0, DenseNet-201, and Inception-V3, in addition to state-of-the-art (SOTA) classifiers. The effectiveness of the proposed model is further demonstrated by the results of the ablation experiments. The proposed model was successful in achieving an accuracy of 99.01%, making it superior to both the baseline models and the SOTA classifiers. As a result, the proposed approach is capable of offering significant support to radiologists and other medical professionals.

High authenticity steering feel control strategy for Steer-by-Wire system based on terminal sliding mode rack force observer

This article proposes a high authenticity steering feel control strategy for Steer-by-Wire (SBW) system based on terminal sliding mode rack force observer, in response to the issue of how to design the steering feel of the SBW system to reflect road conditions. Firstly, a terminal sliding mode rack force observer is designed to achieve real-time and accurate estimation of steering system rack force. Then, based on the vehicle motion state, design the feedback characteristic curve of the rack force and the compensation torque of the handwheel to obtain the expected steering feel, and use the automatic disturbance rejection controller (ADRC) method to control the voltage of the handwheel motor. In simulation testing, a SBW open loop control system model is established to verify the performance of the designed rack force observer. The objective evaluation indicators are used to verify that the designed feedback method can intuitively adjust and meet the driver's steering feel needs, and the impact of various compensation torques on the evaluation indicators is analyzed. Finally, the results of actual vehicle experiments indicate that the designed method can provide suitable steering feel, verifying the feasibility of the design technology.

Classification of optical coherence tomography images using deep machine learning methods

Backgraund. Optical coherence tomography (OCT) is a modern high-tech and informative method for detecting pathology of the retina and preretinal layers of the vitreous body. However, the description and interpretation of the research results require high qualifications and special training of an ophthalmologist, and significant time expenditure for the doctor and the patient. At the same time, the use of mathematical models based on artificial neural networks (ANN- models) currently makes it possible to automate many processes associated with image processing. Therefore, solving problems associated with automating the process of classifying OCT images based on ANN models is actual.&#x0D; Aims. To develop architectures of mathematical (computer) models based on deep learning of convolutional neural networks (CNN) for classification of OCT images of the retina. To compare the results of computational experiments conducted using Python tools in the Google Colaboratory with single-model and multi-model approaches and evaluate classification accuracy. To make conclusions about the optimal architecture of ANN models and the values of the hyperparameters used.&#x0D; Materials and methods. The original dataset, which was anonymized OCT images of real patients, included more than 2000 images obtained directly from the device in a resolution of 1920 × 969 × 24 BPP. The number of image classes is 12. To create the training and validation data sets, a subject area of 1100 × 550 × 24 BPP was “cut out.” Various approaches were studied: the possibility of using pretrained CNNs with transfer learning, techniques for resizing and augmenting images, as well as various combinations of hyperparameters of ANN-models. When compiling the model, the following parameters were used: Adam optimizer, categorical_crossentropy loss function, accuracy metric. All technological processes with images and ANN-models were carried out using Python language tools in Google Colaboratory.&#x0D; Results. Single-model and multi-model principles for classifying OCT images of the retina are proposed. Computational experiments on automated classification of such images obtained from a DRI OCT Triton 3D tomograph using various ANN model architectures showed an accuracy of 98-100% during training and validation and 85% during an additional test, which is a satisfactory result. The optimal architecture of the ANN model - a six-layer convolutional network - was selected and the values of its hyperparameters were determined.&#x0D; Conclusions. The results of deep training of convolutional neural network models with various architectures, their validation and testing showed satisfactory classification accuracy of retinal OCT images. These developments can be used in decision support systems in the field of ophthalmology.

On radially symmetric oscillations of a collisional cold plasma

We study the influence of the friction term on the radially symmetric solutions of the repulsive Euler–Poisson equations with a non‐zero background, corresponding to cold plasma oscillations in multiple spatial dimensions. It is shown that for any arbitrarily small non‐negative constant friction coefficient, there exists a neighborhood of the zero equilibrium in the norm such that the solution of the Cauchy problem with initial data belonging to this neighborhood remains globally smooth in time. Moreover, this solution converges to zero as . This result contrasts with the situation of zero friction, where any small deviation from the zero equilibrium generally leads to a blow‐up. Our method allows us to estimate the lifetime of smooth solutions. Further, we prove that for any initial data, one can find such a coefficient of friction that the respective solution to the Cauchy problem keeps smoothness for all and converges to zero. We also present the results of numerical experiments for physically reasonable situations, which allows us to estimate the value of the friction coefficient, which makes it possible to suppress the formation of singularities of solutions.