«Systems and Means of Informatics»
Scientific journal
Volume 36, Issue 2, 2026
Content | About Authors
Abstract and Keywords
- Yu. A. Stepchenkov
- Yu. G. Diachenko
- D. V. Khilko
- G. A. Orlov
- G. S. Appolonov
- D. Yu. Diachenko
Abstract: The article explores the self-timed (ST) binary counter automated design based on the initial Verilog description of a synchronous prototype. Counters are among the most popular digital units used in robotic systems. Typical synchronous circuit synthesizers construct counters using a storage register and a combinational environment that calculates a new counter state based on its current state and the signal values that determine the counter's properties. However, in most cases, this approach to ST counter synthesis results in excessive hardware redundancy and degraded counter performance. However, circuit design solutions for sequential ST counters are available that ensure minimal hardware costs and acceptable performance. The design features of ST counters with various options for asynchronous and self-timed setup as well as operation-enable features are examined. A method for the ST counter formalized construction based on the use of ready-made hardware blocks - templates - that implement the specific features of their behavior is proposed. The ST counter's Verilog description is assembled from hardware modules corresponding to the specified synthesized counter's options. The proposed template method enables the synthesis of binary ST counters of various types (up, down, and reversible), automates the creation of a cell library for ST circuit synthesis, and guarantees the self-timing of the resulting hardware counter implementations.
Keywords: self-timed circuit; counter; automated design; parameterization; template; cell library
- I. N. Sinitsyn
- V. I. Sinitsyn
- E. R. Korepanov
- T. D. Konashenkova
Abstract: Methodological and algorithmical support for wavelet neural network (WNN) synthesis for observable linear scalar Pugachev stochastic system (StS) with parametric noises (PN) and mean square error (MSE) criterion is presented.
Input of StS with PN contains useful signal scalar stochastic process (StP) depending upon random vector parameters and PN. Output StP is fixed. Linear operator MSE-optimal StS is derived by approximate solution of operator equation connecting second probability moments of input and needed output using methods of multiscale analysis (MSA) and WNN. Input StP is presented in the form of linear combination of input random variables (RV) by means of wavelet canonical expansions (WLCE). Mean-square optimal estimate of an output StP is also constructed in the form of linear combination of input RV with coefficients defined by operator equation solution. Formulae based on the first and second probability moments for accuracy of MSE estimate are given. Computer experiments confirm advantages of WNN synthesis based on MSA, WLCE, and WNN in comparison with recurrent synthesis based on MSA and WLCE.
Keywords: canonical expansion; mean-square criterion; modeling; optimal estimate; parametric noise; stochastic process; stochastic system; wavelet; wavelet-neural network
Abstract: The paper presents a statistical analysis of the equivalence between nonparametric and semiparametric approaches to estimating the random coefficients of an Ito stochastic differential equation which is employed to model turbulent heat fluxes between the ocean and atmosphere. The Wilcoxon-Mann- Whitney test was utilized to evaluate the performance of these estimators as a function of the number of bins used during the input data discretization phase.
The study was conducted using both synthetic data sets with predefined parameters and empirical ERA5 reanalysis data for the North Atlantic. The results indicate that as the number of bins increases, the accuracy of the nonparametric method deteriorates for the drift coefficient but improves for the diffusion coefficient, whereas the semiparametric method demonstrates high stability. Based on statistical testing with the Holm-Bonferroni correction for multiple hypotheses testing, a threshold number of bins (approximately 200-250) was identified, beyond which the distributions of estimates from both methods become statistically indistinguishable. This confirms their practical equivalence for the analysis of geophysical data.
Keywords: Ito stochastic differential equation; random coefficients; EM algorithm
- A. S. Golovin
- E. V. Morozov
- A. S. Rumyantsev
Abstract: An M/G/1-type model with deactivation and activation periods following a hot-standby period in the empty state is considered where all periods possess general distributions. The model is treated as an exceptional first service system and, together with regenerative arguments, it allows one to obtain the stationary performance (workload, response time) distributions (in the transform domain) as well as the average power demand in explicit form. The optimal value of the hot-standby parameter which minimizes the average power demand (under restricted performance degradation) is obtained. It is shown that the constant-time hot-standby policy is sufficient to guarantee optimality. These results are useful to study power and performance tradeoffs in larger models such as a heterogeneous server pool as shown by a numerical study.
Keywords: delayed deactivation; energy efficiency; exceptional first service; performance monotonicity; Laplace transform order; optimal policy
- M. A. Ivanov
- V. Yu. Korolev
Abstract: A class of multivariate elliptically contoured distributions is introduced and studied. Each one-dimensional projection of such a distribution has the quasi-exponentiated normal distribution that coincides with the distribution of the radom variable  Q is the positive random variable; and X is the random variable with the standard normal distribution independent of Q. For  > 1, the densities of the multivariate mixed normal distributions are infinite in zero. This property makes it possible to use multivariate quasi-exponentiated mixed power distributions with > 1 as models of statistical regularities in the behavior of multivariate stochastic processes with rather long periods within which the process either does not change or changes insignificantly, alternate with the periods when variations with rather large jumps are observed. Unlike "pure" quasi-exponentiated distributions, quasi-exponentiated mixed normal distributions possess heavy tails that may be useful in the case of very large jumps of the process under consideration. Some limit theorems are presented on convergence of the distributions of multivariate statistics constructed from samples with random sizes, including random sums, to multivariate quasi-exponentiated mixed normal distributions. As an example, multivariate elliptically contoured quasi-exponentiated logistic distribution is considered.
Keywords: quasi-exponentiated normal distribution; normal scale mixture; elliptically contoured distribution; limit theorem; random sum
Abstract: The criteria of the unit root are widely used in the analysis of the stationarity of a time series. There is a detailed substantiation of the variants of such tests which in mathematical statistics are called the Dickey-Fuller criteria.
The efficacy of the proposed methods for describing the limiting distributions of the statistics used has been confirmed in the course of many studies but they turned out to be unproductive at the finite values of the observation time T. Therefore, it was necessary to turn to the method of statistical trials (MST), with the help of which percentile tables were constructed for individual T values. In addition to the fact that they were clearly insufficient, in fact, for the convenience of using preconstructed statistical tables, the researcher was forced to turn to a simple data model in conjunction with a fixed set of artificially selected competing hypotheses. As a result, the only solution to the problem of applying the unit root criteria for small T in practice is the use of MST. Earlier, the author of this article considered the use of cointegration analysis in the ranking of a set of objects on the basis of a single indicator - the degree of connectivity of the components of the observed multidimensional time series.
The newly constructed statistical tables for T = 8 made it possible to confirm the correctness of the replacement of the unit root criteria due to small T with the stationarity criteria of a general nature. The capabilities of modern processors that allow multitasking as well as the prospects for using the R platform speak in favor of the productivity of simulation in the practice of data analysis.
Keywords: cointegration analysis; testing for a unit root; Dickey-Fuller test; relate coefficient; regional economy; investments; gross regional product; tests for stationarity; statistics with R
Abstract: The paper addresses the problem of generating synthetic images for computer vision systems under limited availability of representative real-world datasets. A hybrid algorithm based on stacking generative adversarial and diffusion models is proposed. The key contribution is the modification of the Diffusion-GAN architecture, in which the forward diffusion process is replaced by the mechanism from Stable Diffusion, combining the computational efficiency of diffusion models with the training stability of adversarial approaches. The algorithm implements a three-stage pipeline: training the modified generative model, generating synthetic images, and postprocessing to improve visual quality. Experimental validation was performed on the vehicle detection and classification task using the Vehicle Classification SGCUM dataset. The results demonstrate that the YOLOv8 model trained exclusively on synthetic data achieves accuracy metrics comparable to those of a model trained on real data, confirming the suitability of the generated data for training deep neural networks.
Keywords: deep neural networks; generative-adversarial models; diffusion models; synthetic data; technical vision systems
- I. A. Saitov
- N. I. Fokin
- Yu. B. Mironov
- B. M. Shabanov
Abstract: The article presents an original approach to modeling the quality of delivery of reference time and frequency signals using the spectral correlation theory of random processes. This and similar problems arise in solving modern design (synthesis, optimization) problems of time-frequency support systems for information infrastructure facilities, including those using alternative sources of reference signals. Implementation of this approach will form the foundation for developing new core elements of a theory for constructing a Unified Time- Frequency Support System for the National Information Infrastructure. The core of such a theory is proposed to be an analytical description of the dependence of time and frequency signal quality on the causes and characteristics of their fluctuations. The importance of these patterns stems from the fact that the potential range of their distribution depends on the discount rate of time and frequency signal quality. This, in turn, determines the key topological characteristics and cost of the time-frequency support system for the information infrastructure.
Keywords: time-frequency support; national information infrastructure; synchronization quality
|
