Search

The technologies of artificial intelligence and cloud computing systems have recently been actively developed and implemented. In this regard, the issue of their joint use, which has been topical for several years, has become more acute. The problem of data privacy preservation in cloud computing acquired the status of critical long before the necessity of their joint use with artificial intelligence, which made it even more complicated. This paper presents an overview of both the artificial intelligence and cloud computing techniques themselves, as well as methods to ensure data privacy. The review considers methods that utilize differentiated privacy; secret sharing schemes; homomorphic encryption; and hybrid methods. The conducted research has shown that each considered method has its pros and cons outlined in the paper, but there is no universal solution. It was found that theoretical models of hybrid methods based on secret sharing schemes and fully homomorphic encryption can significantly improve the confidentiality of data processing using artificial intelligence. [ABSTRACT FROM AUTHOR]

Segmentation of whole slide histological images through the classification of tissue types of small fragments is an extremely relevant task in digital pathology, necessary for the development of methods for automatic analysis of whole slide histological images. The extremely large resolution of such images also makes the task of increasing image resolution relevant, which allows storing images at a reduced resolution and increasing it if necessary. Annotating whole slide images by histologists is complex and time-consuming, so it is important to make the most efficient use of the available data, both labeled and unlabeled. In this paper we propose a novel neural network method to simultaneously solve the problems of super-resolution of histological images from ×20 optical magnification to ×40 and classifying image fragments into tissue types at ×20 magnification. The use of a single encoder as well as the proposed neural network training scheme allows to achieve better results on both tasks compared to existing approaches. The PATH-DT-MSU WSS2v2 dataset presented for the first time in this paper was used for training and testing the method. On the test sample, an accuracy value of 0.971 and a balanced accuracy value of 0.916 were achieved in the classification task on 5 tissue types, for the super-resolution task, values of PSNR = 32.26 and SSIM = 0.89 were achieved. The source code of the proposed method is available at: https://github.com/Kukty/WSI_SR_CL. [ABSTRACT FROM AUTHOR]

Currently, the digitization of environment objects (vegetation, terrain, architectural structures, etc.) in the form of point clouds is actively being developed. The integration of such digitized objects into virtual environment systems allows the quality of the modeled environment to be improved, but requires efficient methods and algorithms for real-time visualization of large point volumes. In this paper the solution of this task on modern multicore GPUs with support of hardware-accelerated ray tracing is researched. A modified method is proposed where the original unordered point cloud is split up into point groups whose visualization is effectively parallelized on ray tracing cores. The paper describes an algorithm for constructing such groups using swapping arrays of point indices, which works faster than alternative solutions based on linked lists, and also has lower memory overhead. The proposed method and algorithm were implemented in point cloud visualization software complex and approbated on a number of digitized environment objects. The results of the approbation confirmed the efficiency of proposed solutions as well as their applicability for virtual environment systems, video simulators, geoinformation systems, virtual laboratories, etc. [ABSTRACT FROM AUTHOR]

Instrument panels of modern aircrafts are created based on the concept of glass cockpit. This new interface philosophy improves the perception of important flight information by displaying it on a single multi-function display. This paper considers problems that arise when developing a certified cockpit display visualization system designed for operation on civil aircrafts under the Russian real-time operating system JetOS. The paper presents several algorithmic solutions that make it possible to achieve acceptable rendering speed. In particular, a solution to the problem of rigid scheduling of operating system partitions is described in detail. This solution makes it possible to overcome the degradation of rendering speed. Some directions for further research are outlined. [ABSTRACT FROM AUTHOR]

This paper considers theoretical foundations and mathematical methods of data analysis under the conditions of the Rice statistical distribution. The problem involves joint estimation of the signal and noise parameters. It is shown that this estimation requires the solution of a complex system of essentially nonlinear equations with two unknown variables, which implies significant computational costs. This study is aimed at mathematical optimization of computer algebra methods for numerical solution of the problem of Rician data analysis. As a result of the optimization, the solution of the system of two nonlinear equations is reduced to the solution of one equation with one unknown variable, which significantly simplifies algorithms for the numerical solution of the problem, reduces the amount of necessary computational resources, and enables the use of advanced methods for parameter estimation in information systems with priority of real-time operation. Results of numerical experiments carried out using Wolfram Mathematica confirm the effectiveness of the developed methods for two-parameter analysis of Rician data. The data analysis methods considered in this paper are useful for solving many scientific and applied problems that involve analysis of data described by the Rice statistical model. [ABSTRACT FROM AUTHOR]

In this paper, the authors present a technique for developing a secure messaging service called "CryptoMess" which utilizes the Commutative Supersingular Isogeny Diffie-Hellman (CSIDH) algorithm for secure key exchange and the Advanced Encryption Standard (AES) to protect message content in communication. In addition, the authors have incorporated the Falcon post-quantum digital signature technology to ensure the integrity and authenticity of communications between the sender and the recipient. The novel post-quantum cryptographic algorithms utilized by the authors are still under consideration in the competition to select cryptosystems for use in quantum computing systems organized by the NIST Standards Institute. As a result, the messaging program "CryptoMess" is able to exchange messages between users, providing safety, security, integrity, and authenticity. The authors have included modern post-quantum cryptography techniques, such as the Falcon digital signature system, to guarantee that the product operates safely. The program has a transmitting rate of approximately 209–261 ms and a receiving rate of approximately 168–206 ms. The message signing time is about 260 ms, and the message verification speed is approximately 185 ms. Key generation time is about 741 ms, key encapsulation time is about 1.454 ms, and key decapsulation time is about 1.921 ms. The source code of the "CryptoMess" communications program has been analyzed and tested by the authors using the Fortify Static Code Analyzer tool to ensure that the product has been created safely and securely. [ABSTRACT FROM AUTHOR]

Grammatical error correction is one of the core natural language processing tasks. Presently, the open-source state-of-the-art sequence tagging for English is the GECToR model. For Russian, this problem does not have equally effective solutions due to the lack of annotated datasets, which motivated the current research. In this paper, we describe the process of creating a synthetic dataset and training the model on it. The GECToR architecture is adapted for the Russian language, and it is called RuGECToR. This architecture is chosen because, unlike the sequence-to-sequence approach, it is easy to interpret and does not require a lot of training data. The aim is to train the model in such a way that it generalizes the morphological properties of the language rather than adapts to a specific training sample. The presented model achieves the quality of 82.5 in the metric on synthetic data and 22.2 on the RULEC dataset, which was not used at the training stage. [ABSTRACT FROM AUTHOR]

This paper presents a comparative efficiency analysis of hashing algorithms from a perspective of their use in systems based on the zero-knowledge succinct non-interactive argument of knowledge (zk-SNARK). Hash functions sha256, sha3, poseidon, mimc, and blake2, most commonly used in modern distributed ledgers, are considered. For parameter measurement experiments, an infrastructure using the ZoKrates toolbox is developed. Based on experimental data, the limits of practical applicability of the algorithms to the problem of proving knowledge of a hash preimage with zk-SNARK circuits in distributed ledgers are determined, and related efficiency problems are considered. [ABSTRACT FROM AUTHOR]

This paper explores the possibility of using dual representation of geometry to speed up ray tracing and ensure robustness of light propagation simulation in complex optical systems with freeform surfaces defined by high-degree polynomials (up to order 34) or Jacobi polynomials. Traditional methods for representing this geometry both as a triangular mesh and as an analytical expression are analyzed. The analysis demonstrates the disadvantages of the traditional approaches due to the insufficient accuracy of calculating coordinates of the point at which the ray intersects the triangular mesh, as well as the non-robustness of the conventional procedures for finding the point of intersection between the tangent ray and the analytical surface. Thus, it is proposed to use a dual representation of geometry as a rough approximation of the surface by a triangular mesh, which is subsequently used as an initial approximation to find the point at which the ray intersects the surface defined by an analytical expression. This approach significantly speeds up convergence of the analytical methods and improves robustness of their solutions. The use of the Intel® Embree library to quickly find the point of intersection between the ray and the rough triangular mesh, as well as a vector model to refine the coordinates of the point of intersection between the ray and the analytically represented geometry, allows us to develop and implement a ray tracing algorithm for an optical system that has surfaces with dual representation of geometry. Experiments carried out using the developed algorithm show a significant speedup of ray tracing while preserving computational accuracy and high robustness of the results. The results are demonstrated by evaluating the point spread function and glare for two lenses with freeform surfaces defined by Jacobi polynomials. In addition, for these two lenses, an image formed by an RGB-D object that simulates a real scene is calculated. [ABSTRACT FROM AUTHOR]

This paper presents an approach to 3D model reconstruction from point cloud data using modern neural network architectures. The method is based on PointNet++ and Transformer. PointNet++ plays a key role, providing efficient feature extraction and encoding of complex geometries in 3D scenes. This is achieved by recursively applying PointNet++ to nested partitions of the input point set in a metric space. Convex decomposition, which is an important step in the proposed approach, transforms complex 3D objects into sets of simpler convex shapes. This facilitates data processing and makes the reconstruction process more manageable. Then, Transformer trains the model on the extracted features, which enables the generation of high-quality reconstructions. It should be noted that Transformer is used only to determine positions of the walls and detect object boundaries. This combination of technologies allows us to achieve high accuracy of 3D model reconstruction. The main idea of the method is to segment the point cloud into small fragments, which are then reconstructed as polygon meshes. To restore missing points in point cloud data, a method based on the L1-median algorithm and local features of the point cloud is used. This approach can adapt to different geometric structures and correct topological connectivity errors. The proposed method is compared with several modern approaches, demonstrating its potential in various fields, including architecture, engineering, digitalization of cultural heritage, as well as augmented and mixed reality systems. This confirms its wide applicability and great perspectives for further development. [ABSTRACT FROM AUTHOR]

In the modern world, digital image processing requires increasing the speed of the processing methods and algorithms used. One way to improve performance is to transform spatial filters into filters with a recursively separable form of implementation. The recursion property implies the use of previous output values of a function to form the current sample. The property of separability is understood as division into processing by column and row of a matrix of digital image values. The transformation of spatial filters consists of changing the aperture of the masks into a non-orthogonal (non-equilateral) form, which reduces the number of computational operations and speeds up the processing process, while maintaining its efficiency. The paper presents a description of the non-equilateral apertures of the previously developed Laplacian digital filters "truncated pyramid" and "double pyramid." For non-equilateral apertures, results were obtained for the first time on their use for television measuring systems. From which it can be seen that a "truncated pyramid" Laplacian filter with non-equilateral processing apertures is recommended for use in TIS, since it increases the efficiency of measuring the range to objects of interest while reducing processing time. Based on the results of processing with modified filters, sets of processed images were obtained for each of the 10 original images. For each set of processed images, measurements of the peak signal-to-noise ratio, standard deviation and selection of the optimal central coefficient of the filter mask were carried out, for subsequent assessment of the effectiveness of processing with modified filters. The assessment of the influence of recursively separable Laplacian filters "truncated pyramid" and "double pyramid" with non-equilateral aperture masks on the accuracy of television measurement systems consisted of considering their influence on measuring the distance (from the camera) to the object of interest in the image, when - control of processing time. Based on the evaluation results, we can conclude that by using pre-processing of images with modified digital filters, the accuracy of measuring the distance from the camera to the measurement object is improved, while reducing processing time. [ABSTRACT FROM AUTHOR]

In this paper we consider the problem of detecting blurred regions in high-resolution whole slide histologic images. The proposed method is based on the use of a Fourier neural operator trained on the results of two simultaneously used approaches: blur detection using multiscale analysis of the discrete cosine transform coefficients and estimation of the degree of sharpness of objects edges in the image. The efficiency of the algorithm is confirmed on images from the datasets PATH-DT-MSU [1] and FocusPath [2]. [ABSTRACT FROM AUTHOR]

In this paper, we continue to consider the problem of recovering the port-Hamiltonian structure for an arbitrary system of differential equations. We complement our previous study on this topic by explaining the choice of machine learning algorithms and discussing some details of their application. We also consider the possibility provided by this approach for a potentially new definition of canonical forms and classification of systems of differential equations. [ABSTRACT FROM AUTHOR]

This paper considers the problem of constructing compartmental models of dynamic systems by using a software package for symbolic calculation written in Julia. The software package is aimed at unifying the formalized construction of compartmental models, taking into account the meaningful description of possible interactions among compartments and the influence of various factors on the evolution of systems. An approach to the development of the instrumental and methodological basis for modeling the dynamic systems the behavior of which can be described by one-step processes is developed. The proposed software package enables the symbolic representation of the differential equations of the model in both stochastic and deterministic cases. It is implemented in Julia and uses the Julia Symbolics computer algebra library. A comparison between the Julia Symbolics tools and some other computer algebra systems is carried out. The application of the developed software package to a compartmental model is considered. The results can be used to solve problems of constructing and studying dynamic models in natural sciences that are represented by one-step processes. [ABSTRACT FROM AUTHOR]

This scientific paper explores the operation of a multiprocessor task servicing system. Tasks are received into the system at random intervals and are characterized by several stochastic parameters, including the number of processors required for their execution, the maximum allowable busy time for these processors, and the permissible waiting time in the task queue. The organization of task servicing in this system follows a first-in, first-out (FIFO) approach, ensuring uninterrupted processing. The key servicing process involves periodically selecting the first task in the queue and assessing its feasibility for immediate execution. If the task meets the necessary criteria, it is dispatched for processing. This process continues iteratively until a task is found, the parameters of which prevent immediate servicing. It is important to note that tasks in the queue have a limited window of time within which they can be serviced; otherwise, they may exit the system without service. This paper focuses on systems characterized by exponential distributions for random variables related to task arrivals, servicing times, and waiting restrictions. A system of equations is derived that describes the system's steady-state behavior. These equations enable the calculation of probabilities associated with the system's various states. Additionally, the paper provides insights into the probability distributions of virtual waiting times for tasks that arrive in the system at any given moment. [ABSTRACT FROM AUTHOR]

This paper presents results of development of a numerical model of Lagrangian particle transport, as well as results of application of parallel computation methods to improve the efficiency of the software implementation of this model. The model is a software package that allows the transport and deposition of aerosol particles to be calculated taking into account properties of particles and the input data that describe atmospheric conditions and underlying surface geometry. The dynamic core, physical parameterizations, numerical implementation, and algorithm of the model are described. Results of successful verification of the model on analytical solutions are presented. Initially, the model was used for less computationally intensive problems. In this paper, given the need to use the model in more computationally intensive problems, we optimize the sequential software implementation of the model, as well as develop its software implementations that use parallel computing technologies (OpenMP, MPI, and CUDA). The results of testing different implementations of the model show that the optimization of the most computationally complex blocks in its sequential version can reduce the execution time by 27%. At the same time, the use of parallel computing technologies allows us to achieve acceleration by several orders of magnitude. The use of OpenMP in the dynamic block of the model provides almost 4-fold acceleration of this block; the use of MPI, almost 8-fold acceleration; and the use of CUDA, almost 16-fold acceleration (all other conditions being equal). We also give some recommendations on the choice of a parallel computing technology depending on the properties of a computing system. [ABSTRACT FROM AUTHOR]

This paper shows quantitative research regarding knowledge, soft and hard skills, and experience acquired by students hired by a University Software Development Company (USDC). Additionally, suggestions regarding how to set up a USDC in an academic environment, facing real customers, are shown. There have been good and bad experiences, both will be presented in this paper. Furthermore, students' perceptions will be discussed. To identify students' perceptions a questionnaire (survey) was applied. Its reliability was calculated through Cronbach's alpha coefficient (α = 89). Additionally, the Pearson correlation coefficient was calculated (r) in order to identify questions that should be deleted to increase the questionnaire's reliability. Outcomes could be useful when a software engineering faculty wishes to set up a USDC. [ABSTRACT FROM AUTHOR]

The field of human event recognition using visual data in smart environments has emerged as a fruitful and successful area of study, with extensive research and development efforts driving significant advancements. These advancements have not only provided valuable insights, but also led to practical applications in various domains. In this context, human actions, activities, interactions, and behaviors can all be considered as events of interest in smart environments. However, when it comes to smart classrooms, there is a lack of unified consensus on the definition of the term "human event." This lack of agreement presents a significant challenge for educators, researchers, and developers, as it hampers their ability to precisely identify and classify the specific situations that are relevant within the educational context. The aim of this paper is to address this challenge by conducting a systematic literature review of relevant events in smart classrooms, with a focus on their applications in assistive technology. The review encompasses a comprehensive analysis of 227 published documents spanning from 2012 to 2022. It delves into key algorithms, methodologies, and applications of vision-based event recognition in smart environments. As the primary outcome, the review identifies the most significant events, classifying them according to single person behavior, or multiple-person interactions, or object-person interactions. It also examines their practical applications within the educational context. The paper concludes with a discussion on the relevance and practicality of vision-based human event recognition in smart classrooms, particularly in the post-COVID era. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the effectiveness of classical approaches to active learning in the problem of document segmentation with the aim of reducing the size of the training sample. A modified approach to selection of document images for labeling and subsequent model training is presented. The results of active learning are compared to those of transfer learning on fully labeled data. The paper also investigates how the problem domain of a training set, on which a model is initialized for transfer learning, affects the subsequent uptraining of the model. [ABSTRACT FROM AUTHOR]

This paper describes a static analysis tool for finding defects, analyzing metrics and relations for programs written in the Kotlin language. The approach is implemented in the Svace static analyzer developed at the Ivannikov Institute for System Programming of the Russian Academy of Sciences. The paper focuses on the problems we faced during the implementation, the approaches we used to solve them, and the experimental results for the tool we built. The tool not only supports Kotlin but is also capable of analyzing mixed projects that use both Java and Kotlin. We hope that this paper will be useful to static analysis developers and language designers. [ABSTRACT FROM AUTHOR]

This paper discusses the use of a fully connected neural network to classify images of firing pin impressions. The purpose of this work is to investigate the effectiveness of clone images of firing pin impressions in improving the quality of training of fully connected neural networks. Another purpose of the work is to estimate the accuracy of multigroup classification of firing pin impressions left by different firearms by using a neural network. The scientific novelty of this work is in the use of augmentation for creating images of firing pin impressions to increase the number of objects in the training dataset and to artificially improve the feature diversity of objects of each class. The conducted investigation shows that the accuracy of classification of the analyzed objects reaches approximately 84% for a fixed value of the classification criterion and 94–98% when the classification is carried out based on three maximum signals on output neurons. The work is of interest to developers of automated ballistic identification systems. [ABSTRACT FROM AUTHOR]

While modern rendering systems are efficient for modeling complex light paths in complex environments, the rendering of refractive caustics still takes a long time. Caustics are light patterns that occur when light is refracted and reflected from a surface. Due to an irregular density distribution of these specular events, rendering algorithms mainly rely on direct sampling of the bidirectional dispersion distribution function on these surfaces to plot trajectories. This requires a lot of calculations. Photonic maps are also used. However, there are difficulties that limit the applicability of caustic maps. Since each photon in the photon buffer must be processed, one has to choose between a strongly underestimated caustic sampling and a large decrease in speed in order to use a sufficient number of photons for caustics in order to obtain high-quality images. Complex specular interactions cause oversampling in bright focal areas, while other areas of the caustic map remain undersampled and noisy. At the same time, speed takes precedence over realism in most interactive applications. However, the desire to improve the quality of graphics prompted the development of various fast approximations for realistic lighting. This paper presents a combined method for rendering refraction of light and caustics using backward integration for illumination and direct integration for viewing rays. An approach for simultaneous propagation of light and for tracking rays in a volume is used; therefore, it does not require storing data of an intermediate volume of illumination. In the implementation of this method, the distance between the light planes is set to one voxel, which provides at least one sample per voxel for all orientations. The method does not use preliminary calculations, and all rendering parameters can be changed interactively. As a result, using the proposed method, it is possible to create plausible approximations of complex phenomena, such as refractions and caustics. The effect of refraction on the shadow is shown. Complex light patterns occurring due to the curved geometry of objects are demonstrated. The visualization results show the importance of refraction for the appearance of transparent objects, e.g., the effect distortions caused by refraction in the medium and refraction on the interfaces between media. The difference in refractive indices between different media causes a complex interaction between light and shadow areas. It is shown how refraction and caustics improve rendering of functionally defined objects by providing additional information about their shape and location. [ABSTRACT FROM AUTHOR]

This paper presents an algorithm and description of its software implementation for detection of lineaments (ground erosions or cracks) in aerial images of open pits. The proposed approach is based on the apparatus of convolutional neural networks for semantic classification of binarized images of lineament objects, as well as graph theory for determining the geometric location of linearized lineament objects with subsequent calculation of their lengths and areas. As source data, three-channel RGB images of high-resolution aerial photography (10×10 cm) are used. The software module of the model is logically divided into three levels: preprocessing, detection, and post-processing. The first level implements the preprocessing of input data to form a training sample based on successive transformations of RGB images into binary images by using the OpenCV library. A neural network of the U-Net type, which includes convolutional (Encoder) and scanning (Decoder) blocks, represents the second level of the information model. At this level, automatic detection of objects is implemented. The third level of the model is responsible for calculating their areas and lengths. The result provided by the convolutional neural network is passed to it as input data. The lineament area is calculated by summing the total number of points and multiplying by the pixel size. The lineament length is calculated by linearizing the areal object into a segmented object with node pixels and, then, calculating the lengths between them while taking into account the resolution of the source image. The software module can work with fragments of the source image by combining them. The module is implemented in Python and its source code is available at https://gitlab.ict.sbras.ru/popov/lineaments/-/tree/master/lineaments-cnn. [ABSTRACT FROM AUTHOR]

In this paper, based on height field surface example, an efficient technology of real-time modeling of complex procedural objects on the ray tracing pipeline (RT-pipeline) is proposed. The proposed technology doesn't overload the I-shader stage (intersection shader), but distributes the computational load between the I-shader and the AH-shader (any-hit shader). The key innovations of the technology are the early rejection at the I-shader stage of the bounding boxes (AABBs) extracted by the RT-pipeline hardware unit, and the "transparent AABB" concept which allows transferring costly computing of the "ray-procedural object" intersection to a later AH-shader stage. The paper also describes a number of modifications that reduce the amount of such calculations. The proposed technology was implemented in a software complex in C++, GLSL and using the Vulkan API. The performance of the developed solution was studied under various ray tracing conditions on the task of modeling the surface of a detailed Puget Sound height field. The obtained results confirmed efficiency of the developed technology and the possibility of its application in virtual environment systems, simulators, scientific visualization, etc. [ABSTRACT FROM AUTHOR]

In this paper, we consider the problem of augmenting a set of histological images with adversarial examples to improve the robustness of neural network classifiers trained on the augmented set against adversarial attacks. In recent years, neural network methods have been developing rapidly, achieving impressive results. However, they are vulnerable to so-called adversarial attacks; i.e., they make incorrect predictions on input images with added carefully crafted imperceptible noise. Hence, the reliability of neural network methods remains an important area of research. In this paper, we compare different methods for training set augmentation to improve the robustness of neural histological image classifiers against adversarial attacks. For this purpose, we augment the training set with adversarial examples generated by several popular methods. [ABSTRACT FROM AUTHOR]

This paper discusses the problem of classifying images of land impressions on discharged bullets in terms of the "match" and "non-match" categories. The research is aimed at improving the effectiveness of comparing land impression images by the congruent matching profile segments (CMPS) method. The scientific novelty of the approach is in supplementing the analysis with an additional independent feature, as well as in using the k-nearest neighbors algorithm at the final stage of trace comparison. The research shows that the accuracy of classification of the compared pairs of land impression images by the combined method is approximately 87%. The analysis by the CMPS method makes it possible to effectively compare land impression images with high resolution (approximately 1 μm per pixel). The research is of interest to developers of automated ballistic identification systems. [ABSTRACT FROM AUTHOR]

There is a wide range of fuzzing techniques for security assurance. Communication protocols and programs accepting complex structured data, including compilers, can be efficiently tested with a corresponding fuzzing approach. The main advantage of the fuzzing is actual data, which demonstrates the defects of the target software. The performance of fuzzing primarily depends on the quality of generated data. Existing fuzzing tools provide a fixed set of data generation algorithms, which doesn't consider the specifics of the target program. There are several scenarios when the performance of the fuzzing may suffer. The target may execute different parts of the code based on a comparison of the offsets of the input data with some constant value. For such cases, random mutations will invalidate input data and lead to lower code coverage. In this paper, we present a new approach for effective input data generation. We use static analysis to extract information regarding the input data offsets' comparisons with constant values. Then we use this information during input data generation. Experimental evaluation of the developed method on several projects proves the efficiency of the developed method. Our approach allowed up to 91% increase in the number of executed paths for the same execution count. [ABSTRACT FROM AUTHOR]

Various neuron approximations can be used to reduce the computational complexity of neural networks. One such approximation based on summation and maximum operations is a bipolar morphological neuron. This paper presents an improved structure of the bipolar morphological neuron that enhances its computational efficiency and a new approach to training based on continuous approximations of the maximum and knowledge distillation. Experiments were carried out on the MNIST dataset using a LeNet-like neural network architecture and on the CIFAR10 dataset using a ResNet-22 model architecture. The proposed training method achieves 99.45% classification accuracy on the LeNet-like model (the same accuracy as that provided by the classical network) and 86.69% accuracy on the ResNet-22 model compared with 86.43% accuracy of the classical model. The results show that the proposed method with log-sum-exp (LSE) approximation of the maximum and layer-by-layer knowledge distillation makes it possible to obtain a simplified bipolar morphological network that is not inferior to the classical networks. [ABSTRACT FROM AUTHOR]

One of the open problems in modern control theory is synthesis of adaptive control for nonlinear systems with parametric uncertainty and analysis of stability of the corresponding closed-loop system. Fuzzy logic is one of the approaches that can take into account nonlinearity and uncertainty of the plant. Affine systems constitute a class of nonlinear systems often used to solve various practical problems. For this class, there are a number of methods for synthesis of controllers, in particular, a method based on the matrix Riccati equation with state-dependent coefficients. In this paper, for a given class of nonlinear systems, an adaptation mechanism of a neuro-fuzzy controller that approximates the control synthesized by the SDRE method is used for the first time. The main contribution of this work is the architecture of the neuro-fuzzy network and methods for its initialization. The proposed approach is applied to the model of a two-link manipulator with uncertain coefficients. The conducted numerical experiments demonstrate the effectiveness of the synthesized control in terms of the selected quality criteria. [ABSTRACT FROM AUTHOR]

The key feature of optically variable elements is the appearance and disappearance of high saturation colors during shooting under different angles. The accuracy of color saturation measurement algorithms is limited by the automatic white balance correction subsystem, which uncontrollably changes the color characteristics of the image from frame to frame. In this paper, the possibility of using a priori information about color consistency of the document area to compensate for distortion of the white balance is considered using examples of document owner monochrome photographs. The proposed method of color correction increases the contrast of color saturation between the unprotected areas of the photograph and the areas covered by visually variable image on the protective film. The method is tested on reference data obtained with a mobile device. [ABSTRACT FROM AUTHOR]

Code analysis for defect detection is very important in the modern world, especially in the case of complex multi-thread applications. An example of a tool, suitable for software of high complexity, is the famous S2E, which allows for full-system emulation with symbolic execution. This paper presents several major enhancements for S2E, including: firstly, support for multiple virtual cores, allowing to have parallel speed-up; secondly, on this basis, a race checker plugin to detect defects of this sort in multi-thread programs. This development has concerned such interesting points of research as scheduling in multi-core emulation and race detection with symbolic execution. [ABSTRACT FROM AUTHOR]

To better extract the features from text instances with various shapes, a scene text detector using High Resolution Net (HRNet) and spatial attention mechanism is proposed in this paper. Specifically, we use HRNetv2-W18 as the backbone network to extract the text feature in text instances with complex shapes. Considering that the scene text instance is usually small, to avoid too small feature size, we optimize HRNet through deformable convolution and Smooth Maximum Unit (SMU) activation function, so that the network can retain more detail information and location information of the text instance. In addition, a Text Region Attention Module (TRAM) is added after the backbone to make it pay more attention to the text location information and a loss function is used to TRAM, so that the network can learn the features better. The experimental results illustrate that the proposed method can compete with the state-of-the-art methods. Code is available at: https://github.com/zhangyan1005/HR-DBNet. [ABSTRACT FROM AUTHOR]

The convergence rate has been widely accepted as a performance measure for choosing a better metaheuristic algorithm. So, we propose a novel technique to improve the performance of the existing Grey Wolf Optimization (GWO) algorithm in terms of its convergence rate. The proposed approach also prioritizes the test cases that are obtained after executing the input benchmark programs. This paper has three technical contributions. In our first contribution, we generate test cases for the input benchmark programs. Our second contribution prioritizes test cases using an improved version of the existing GWO algorithm (CGWO). Our third contribution analyzes the obtained result and compares it with state-of-the-art metaheuristic techniques. This work is validated after running the proposed model on six benchmark programs. The obtained results show that our proposed approach has achieved 48% better APFD score for the prioritized order of test cases than the non-prioritized order. We also achieved a better convergence rate, which takes around 4000 fewer iterations, when compared with the existing methods on the same platform. [ABSTRACT FROM AUTHOR]

This paper presents an extension for the Practice Library of the Essence language in the form of a practice for working with antipatterns. To represent antipatterns in a system, the Antipattern subalpha, its states, and checkpoints are proposed. To record data about an antipattern, the Antipattern Report work product and its levels of detail with checkpoints are proposed. To analyze the architecture of a system, the Inspect Architecture activity is proposed. The Fix Architecture activity represents actions for fixing architecture flaws. Code analysis is represented as the Review the Code activity, while the correction of deficiencies found during the analysis is represented as the Refactor the Code activity. The effect of the Antipattern subalpha on the state of the Software System alpha is analyzed. Some recommendations concerning the proposed activities are provided. [ABSTRACT FROM AUTHOR]

The widespread digitalization of the modern society and the development of information technology have increased the number of ways in which financial institutions and potential consumers of financial services can interact. At the same time, the advent of new financial products inevitably leads to a rise in threats, and the use of information technology facilitates the constant "improvement" of fraud schemes and unfair business practices, which negatively affect both the financial market as a whole and its individual participants such as financial institutions and their clients. With the development of the modern society, most financial transactions, including the fraudulent ones, have moved to the Internet. When services are provided remotely, it is more difficult to trace and prosecute the beneficiary. However, there are still ways to stop fraudulent activity, but they are associated with high costs of monitoring and analysis of huge amounts of unstructured information (BigData) available on the Internet, which takes a great amount of time and effort. A solution to illegal activity detection in financial markets is based on open data intelligence, machine learning, and systems analysis. This paper examines certain types of financial services provided on the Internet among which fraudulent activities are most common. In order to identify illegal financial services, some criteria are developed and grouped based on their contribution to the decision-making process. The main result of this study is the construction of a scale for a complex indicator, which is used to build a mathematical model based on the developed criteria and machine learning methods for determining the degree of illegality of online financial services. [ABSTRACT FROM AUTHOR]

This paper presents results of computational and experimental studies of the evolution of turbulent mixing in three-layer gas systems with the development of hydrodynamic instabilities, in particular, the Richtmyer–Meshkov and Kelvin–Helmholtz instabilities, under the action of shock waves. One of the contact boundaries between gases is flat, while the other one has the form of a chevron. The numerical simulations are carried out both with and without initial perturbations of contact boundaries. It is shown that the roughness of the contact boundary significantly affects the width of the mixing zone. [ABSTRACT FROM AUTHOR]

The study of formal stability of equilibrium positions of a multiparametric Hamiltonian system in a generic case is traditionally carried out using its normal form under the condition of the absence of resonances of small orders. In this paper we propose a method of symbolic computation of the condition of existence of a resonance of arbitrary order for a system with three degrees of freedom. It is shown that this condition for each resonant vector can be represented as a rational algebraic curve. By methods of computer algebra the rational parametrization of this curve for the case of an arbitrary resonance is obtained. A model example of some two-parameter system of pendulum type is considered. [ABSTRACT FROM AUTHOR]

Any software development project needs to estimate non-functional requirements (NFR). Typically, software managers are forced to use expert judgment to estimate the NFR. Today, NFRs cannot be measured, as there is no standardized unit of measurement for them. Consequently, most estimation models focus on the functional user requirements (FUR) and do not consider the NFR in the estimation process because these terms are often subjective. The objective of this paper was to show how an NFR estimation model was created using fuzzy logic, and K-Nearest Neighbors classifier algorithm, aiming to consider the subjectivity embedded in NFR terms to solve a specific problem in a Mexican company. The proposed model was developed using a database with real projects from a Mexican company in the private sector. The results were beneficial and better than the initial model considering quality criteria like mean magnitude of relative error (MMRE), standard deviation of magnitude of relative error (SDMRE) and prediction level (Pred 25%). Additionally, the proposed approach allows the managers to identify quantitative elements related to NFR that could be used to interpret the data and build additional models. [ABSTRACT FROM AUTHOR]

Quantum computing is based on the principles of quantum mechanics, such as superposition, entanglement, measurement, and decoherence. The basic units of computation are qubits, which are abstract objects with a mathematical expression to implement the quantum mechanics principles. Alongside quantum hardware, software is a principal element for conducting quantum computing. The software consists of logic gates and quantum circuits that implement algorithms for the execution of quantum programs. Due to those characteristics, quantum computing is a paradigm that non-physics experts cannot understand. Under this new scheme for developing software, it is important to integrate a conceptual framework of the fundamentals on which quantum computing is based. In this paper, we present a kind of taxonomical view of the fundamental concepts of quantum computing and the derived concepts that integrate the emerging discipline of quantum software engineering. We performed a quasi-systematic mapping for conducting the systematic review because the objective of the review only intends to detect the fundamental concepts of quantum computing and quantum software. The results can help computer science students and professors as a starting point to address the study of this discipline. [ABSTRACT FROM AUTHOR]

When a person suffers a stroke, they require rehabilitation to recover from the consequences caused by this event. In order to carry out the rehabilitation, a multidisciplinary team of specialists intervenes, providing care from diverse areas such as neurology, nutrition, psychology, and physiotherapy. In the rehabilitation process, physicians interact with medical computing software and devices. The interactions represent medical activities that follow rehabilitation. However, there is no clear conception of how specialists coordinate to carry out medical activities. There are no specific communication channels that allow multidisciplinary collaboration for complete rehabilitation. This paper presents a systematic review of the state of the art that addresses this problem. On the other hand, a collaborative software architecture is proposed to support this area, which allows the monitoring medical activities from human-computer multimodal interactions. The architecture is composed of 3 layers: (1) to perceive the interactions and to monitor the activities, (2) to control the multidisciplinary access and share the information, and (3) to analyze and evaluate the execution of the multidisciplinary activities. Evaluating how the activities are carried out will help the physicians make decisions regarding the execution of the treatment plan. Therefore, we propose an activities representation diagram that facilitates this evaluation. Finally, we developed a prototype with a user-centered design that perceives human-computer interactions supported by the architecture. [ABSTRACT FROM AUTHOR]

This work is devoted to improving the quality of segmentation of images of various scientific papers and legal acts by neural network models by training them using modified loss functions that take into account special features of images of the appropriate subject domain. The analysis of existing loss functions is carried out, and new functions are proposed that work both with the coordinates of bounding boxes and use information about the pixels of the input image. To assess the quality, a neural network segmentation model with modified loss functions is trained, and a theoretical assessment is carried out using a simulation experiment showing the convergence rate and segmentation error. As a result of the study, rapidly converging loss functions are created that improve the quality of document image segmentation using additional information about the input data. [ABSTRACT FROM AUTHOR]

Because of high complexity of a modern operating system (OS) one has to use complex models in high-level languages even when specifying some separate aspects of OS functionality, such as, security functions. The problem of verifying the conformity between an OS and these models requires the establishment of a rather complex relationship between elements of OS implementation and elements of the model. In this paper, we present a method to establish and support this relationship, which can be effectively used in OS testing and monitoring. This method was successfully used in testing and monitoring the conformity between Linux kernel components and Event-B models. [ABSTRACT FROM AUTHOR]

In this paper, we consider a generalization of the decentralized Frank–Wolfe algorithm to time-varying networks, investigate the convergence properties of the algorithm, and carry out the corresponding numerical experiments. The time-varying network is modeled as a deterministic or stochastic sequence of graphs. [ABSTRACT FROM AUTHOR]

This paper is devoted to some adaptive methods for variational inequalities with relatively smooth and relatively strongly monotone operators. Based on the recently proposed proximal version of the extragradient method for this class of problems, we study in detail the method with adaptively selected parameter values. The rate of convergence of this method is estimated. The result is generalized to the class of variational inequalities with relatively strongly monotone δ-generalized smooth operators. For the ridge regression problem and variational inequality associated with box-simplex games, numerical experiments are carried out to demonstrate the effectiveness of the proposed technique for adaptive parameter selection during the execution of the algorithm. [ABSTRACT FROM AUTHOR]

In this paper, we investigate a variant of the Frank–Wolfe method for convex optimization problems with the adaptive selection of the step parameter corresponding to information about the smoothness of the objective function (the Lipschitz constant of the gradient). Theoretical estimates of the quality of the approximate solution provided by the method using adaptively selected parameters Lk are presented. For a class of problems on a convex feasible set with a convex objective function, the guaranteed convergence rate of the proposed method is sublinear. A special subclass of these problems (an objective function with the gradient dominance condition) is considered and the convergence rate of the method using adaptively selected parameters Lk is estimated. An important feature of the result obtained is the elaboration of the case where it is possible to guarantee, after the completion of the iteration, at least double reduction in the residual of the function. At the same time, the use of adaptively selected parameters in theoretical estimates makes the method applicable to both smooth and non-smooth problems, provided that the iteration termination criterion is met. For smooth problems, it can be proved that the theoretical estimates of the method are reliably optimal up to multiplication by a constant factor. Computational experiments are carried out and a comparison with two other algorithms is made to demonstrate the efficiency of the algorithm on a number of both smooth and non-smooth problems. [ABSTRACT FROM AUTHOR]

This paper focuses on solving a subclass of stochastic nonconvex-nonconcave black box optimization problems with a saddle point that satisfy the Polyak–Łojasiewicz (PL) condition. To solve this problem, we provide the first (to our best knowledge) gradient-free algorithm. The proposed approach is based on applying a gradient approximation (kernel approximation) to an oracle-biased stochastic gradient descent algorithm. We present theoretical estimates that guarantee its global linear rate of convergence to the desired accuracy. The theoretical results are checked on a model example by comparison with an algorithm using Gaussian approximation. [ABSTRACT FROM AUTHOR]

The second smallest eigenvalue of the Laplacian is known as the algebraic connectivity of a graph. It shows degree of graph connectivity. However, this metric does not take into account possible changes in the graph. The removal of even one node or edge can make it disconnected. This work is devoted to the development of a metric that should describe robustness of a graph to such changes. All proposed metrics are based on the algebraic connectivity. In addition, we generalize some well-known optimization methods for our robust modifications of the algebraic connectivity. The paper also reports results of some numerical experiments demonstrating the efficiency of the proposed approaches. [ABSTRACT FROM AUTHOR]

Mathematical formulation of Bohr's complementarity principle leads to the concepts of mutually unbiased bases in Hilbert spaces and complementary quantum observables. In this paper, we consider algebraic structures associated with these concepts and their applications to constructive quantum mechanics. We also briefly discuss some computer-algebraic approaches to the problems under consideration and propose an algorithm for solving one of them. [ABSTRACT FROM AUTHOR]

This paper presents an original package for investigating numerical solutions of ordinary differential equations, which is built in the Sage computer algebra system. This project is focused on a closer integration of numerical and symbolic methods while primarily aiming to create a convenient tool for working with numerical solutions in Sage. The package defines two new classes: initial problems and approximate solutions. The first class defines tools for symbolic computations related to initial problems, while the second class defines tools for interpolating values of symbolic expressions on an approximate solution and estimating the error with the use of the Richardson method. An implementation of the Runge–Kutta method is briefly described, with its main feature being the possibility of working with arbitrary Butcher tableaux and arbitrary numeric fields. [ABSTRACT FROM AUTHOR]

This paper describes the second part of our study devoted to automated anonymization of personal data. The overview and analysis of research prospects is supplemented with a practical result. An anonymization model is proposed, which reduces anonymization of personal data to manipulation of samples of random elements of different types. The key idea of our approach to anonymization of personal data with preservation of their usefulness is the use of the synthesis method, i.e., the complete replacement of all non-anonymized data with synthetic values. In the proposed model, a set of element types is selected, for which corresponding synthesys templates are proposed. The set of templates constitutes a synthesis-based anonymization algorithm. Technically, each template is based on a well-known statistical tool: frequency estimates of probabilities, Parzen–Rosenblatt kernel density estimates, statistical means, and covariances. The proposed approach is illustrated by a simple example from civil aviation. [ABSTRACT FROM AUTHOR]