All issues

The article considers minimization of the expectation of convex function. Problems of this type often arise in machine learning and a variety of other applications. In practice, stochastic gradient descent (SGD) and similar procedures are usually used to solve such problems. We propose to use the ellipsoid method with mini-batching, which converges linearly and can be more efficient than SGD for a class of problems. This is verified by our experiments, which are publicly available. The algorithm does not require neither smoothness nor strong convexity of the objective to achieve linear convergence. Thus, its complexity does not depend on the conditional number of the problem. We prove that the method arrives at an approximate solution with given probability when using mini-batches of size proportional to the desired accuracy to the power −2. This enables efficient parallel execution of the algorithm, whereas possibilities for batch parallelization of SGD are rather limited. Despite fast convergence, ellipsoid method can result in a greater total number of calls to oracle than SGD, which works decently with small batches. Complexity is quadratic in dimension of the problem, hence the method is suitable for relatively small dimensionalities.

The paper studies a multidimensional convection-diffusion equation with variable coefficients and a nonclassical boundary condition. Two cases are considered: in the first case, the first boundary condition contains the integral of the unknown function with respect to the integration variable $x_\alpha^{}$, and in the second case, the integral of the unknown function with respect to the integration variable $\tau$, denoting the memory effect. Similar problems arise when studying the transport of impurities along the riverbed. For an approximate solution of the problem posed, a locally one-dimensional difference scheme by A.A. Samarskii with order of approximation $O(h^2+\tau)$. In view of the fact that the equation contains the first derivative of the unknown function with respect to the spatial variable $x_\alpha^{}$, the wellknown method proposed by A.A. Samarskii in constructing a monotonic scheme of the second order of accuracy in $h_\alpha^{}$ for a general parabolic type equation containing one-sided derivatives taking into account the sign of $r_\alpha^{}(x,t)$. To increase the boundary conditions of the third kind to the second order of accuracy in $h_\alpha^{}$, we used the equation, on the assumption that it is also valid at the boundaries. The study of the uniqueness and stability of the solution was carried out using the method of energy inequalities. A priori estimates are obtained for the solution of the difference problem in the $L_2^{}$-norm, which implies the uniqueness of the solution, the continuous and uniform dependence of the solution of the difference problem on the input data, and the convergence of the solution of the locally onedimensional difference scheme to the solution of the original differential problem in the $L_2^{}$-norm with speed equal to the order of approximation of the difference scheme. For a two-dimensional problem, a numerical solution algorithm is constructed.

Non-destructive acoustic emission testing is an effective and cost-efficient way to examine pressure vessels for hidden defects (cracks, laminations etc.), as well as the only method that is sensitive to developing defects. The sound velocity in the test object and its adequate definition in the location scheme are of paramount importance for the accurate detection of the acoustic emission source. The acoustic emission data processing method proposed herein comprises a set of numerical methods and allows defining the source coordinates and the most probable velocity for each signal. The method includes pre-filtering of data by amplitude, by time differences, elimination of electromagnetic interference. Further, a set of numerical methods is applied to them to solve the system of nonlinear equations, in particular, the Newton – Kantorovich method and the general iterative process. The velocity of a signal from one source is assumed as a constant in all directions. As the initial approximation is taken the center of gravity of the triangle formed by the first three sensors that registered the signal. The method developed has an important practical application, and the paper provides an example of its approbation in the calibration of an acoustic emission system at a production facility (hydrocarbon gas purification absorber). Criteria for prefiltering of data are described. The obtained locations are in good agreement with the signal generation sources, and the velocities even reflect the Rayleigh-Lamb division of acoustic waves due to the different signal source distances from the sensors. The article contains the dependency graph of the average signal velocity against the distance from its source to the nearest sensor. The main advantage of the method developed is its ability to detect the location of different velocity signals within a single test. This allows to increase the degree of freedom in the calculations, and thereby increase their accuracy.

A variant of the inverse method of characteristics (IMH) is presented, in whose algorithm an additional fractional time step is introduced, which makes it possible to increase the accuracy of calculations due to a more accurate approximation of the characteristics. The calculation formulas of the modified method for the equations of the one-velocity model of a gas-liquid mixture are given, with the help of which one-dimensional and also flat test problems with self-similar solutions are calculated. When solving multidimensional problems, the original system of equations is split into a number of one-dimensional subsystems, for the calculation of which the inverse method of characteristics with a fractional time step is used. Using the proposed method, the following were calculated: the one-dimensional problem of the decay of an arbitrary discontinuity in a dispersed medium; a twodimensional problem of the interaction of a homogeneous gas-liquid flow with an obstacle with an attached shock wave, as well as a flow with a centered rarefaction wave. The results of numerical calculations of these problems are compared with self-similar solutions and their satisfactory agreement is noted. On the example of the Riemann problem with a shock wave, a comparison is made with a number of conservative, non-conservative, first and higher orders of accuracy schemes, from which, in particular, it follows that the presented calculation method, i. e. MIMC, quite competitive. Despite the fact that the application of MIMC requires many times more time than the original inverse method of characteristics (IMC), calculations can be carried out with an increased time step and, in some cases, more accurate results can be obtained. It is noted that the method with a fractional time step has advantages over the IMC in cases where the characteristics of the system are significantly curvilinear. For this reason, the use of MIMC, for example, for the Euler equations is inappropriate, since for the latter the characteristics within the time step differ little from straight lines.

Gradient-free optimization methods or zeroth-order methods are widely used in training neural networks, reinforcement learning, as well as in industrial tasks where only the values of a function at a point are available (working with non-analytical functions). In particular, the method of error back propagation in PyTorch works exactly on this principle. There is a well-known fact that computer calculations use heuristics of floating-point numbers, and because of this, the problem of finiteness of the mantissa arises.

In this paper, firstly, we reviewed the most popular methods of gradient approximation: Finite forward/central difference (FFD/FCD), Forward/Central wise component (FWC/CWC), Forward/Central randomization on $l_2$ sphere (FSSG2/CFFG2); secondly, we described current theoretical representations of the noise introduced by the inaccuracy of calculating the function at a point: adversarial noise, random noise; thirdly, we conducted a series of experiments on frequently encountered classes of problems, such as quadratic problem, logistic regression, SVM, to try to determine whether the real nature of machine noise corresponds to the existing theory. It turned out that in reality (at least for those classes of problems that were considered in this paper), machine noise turned out to be something between adversarial noise and random, and therefore the current theory about the influence of the mantissa limb on the search for the optimum in gradient-free optimization problems requires some adjustment.

A nonlinear oscillatory system described by ordinary differential equations with variable coefficients is considered, in which terms that are linearly dependent on coordinates, velocities and accelerations are explicitly distinguished; nonlinear terms are written as implicit functions of these variables. For the numerical solution of the initial problem described by such a system of differential equations, the one-step Galerkin method is used. At the integration step, unknown functions are represented as a sum of linear functions satisfying the initial conditions and several given correction functions in the form of polynomials of the second and higher degrees with unknown coefficients. The differential equations at the step are satisfied approximately by the Galerkin method on a system of corrective functions. Algebraic equations with nonlinear terms are obtained, which are solved by iteration at each step. From the solution at the end of each step, the initial conditions for the next step are determined.

The corrective functions are taken the same for all steps. In general, 4 or 5 correction functions are used for calculations over long time intervals: in the first set — basic power functions from the 2nd to the 4th or 5th degrees; in the second set — orthogonal power polynomials formed from basic functions; in the third set — special linear-independent polynomials with finite conditions that simplify the “docking” of solutions in the following steps.

Using two examples of calculating nonlinear oscillations of systems with one and two degrees of freedom, numerical studies of the accuracy of the numerical solution of initial problems at various time intervals using the Galerkin method using the specified sets of power-law correction functions are performed. The results obtained by the Galerkin method and the Adams and Runge –Kutta methods of the fourth order are compared. It is shown that the Galerkin method can obtain reliable results at significantly longer time intervals than the Adams and Runge – Kutta methods.

The paper develops a new mathematical method of the joint signal and noise parameters determination at the Rice statistical distribution by method of moments based upon the analysis of data for the 1-st and the 3-rd raw moments of the random rician value. The explicit equations’ system have been obtained for required parameters of the signal and noise. In the limiting case of the small value of the signal-to-noise ratio the analytical formulas have been derived that allow calculating the required parameters without the necessity of solving the equations numerically. The technique having been elaborated in the paper ensures an efficient separation of the informative and noise components of the data to be analyzed without any a-priori restrictions, just based upon the processing of the results of the signal’s sampled measurements. The task is meaningful for the purposes of the rician data processing, in particular in the systems of magnetic-resonance visualization, in ultrasound visualization systems, at the optical signals’ analysis in range measuring systems, in radio location, etc. The results of the investigation have shown that the two parameter task solution of the proposed technique does not lead to the increase in demanded volume of computing resources compared with the one parameter task being solved in approximation that the second parameter of the task is known a-priori There are provided the results of the elaborated technique’s computer simulation. The results of the signal and noise parameters’ numerical calculation have confirmed the efficiency of the elaborated technique. There has been conducted the comparison of the accuracy of the sought-for parameters estimation by the technique having been developed in this paper and by the previously elaborated method of moments based upon processing the measured data for lower even moments of the signal to be analyzed.

This article studies a method of constructing a predictive neural network model of a time series based on determining the composition of input variables, constructing a training sample and training itself using the back propagation method. Traditional methods of constructing predictive models of the time series are: the autoregressive model, the moving average model or the autoregressive model — the moving average allows us to approximate the time series by a linear dependence of the current value of the output variable on a number of its previous values. Such a limitation as linearity of dependence leads to significant errors in forecasting.

Mining Technologies using neural network modeling make it possible to approximate the time series by a nonlinear dependence. Moreover, the process of constructing of a neural network model (determining the composition of input variables, the number of layers and the number of neurons in the layers, choosing the activation functions of neurons, determining the optimal values of the neuron link weights) allows us to obtain a predictive model in the form of an analytical nonlinear dependence.

The determination of the composition of input variables of neural network models is one of the key points in the construction of neural network models in various application areas that affect its adequacy. The composition of the input variables is traditionally selected from some physical considerations or by the selection method. In this work it is proposed to use the behavior of the autocorrelation and private autocorrelation functions for the task of determining the composition of the input variables of the predictive neural network model of the time series.

In this work is proposed a method for determining the composition of input variables of neural network models for stationary and non-stationary time series, based on the construction and analysis of autocorrelation functions. Based on the proposed method in the Python programming environment are developed an algorithm and a program, determining the composition of the input variables of the predictive neural network model — the perceptron, as well as building the model itself. The proposed method was experimentally tested using the example of constructing a predictive neural network model of a time series that reflects energy consumption in different regions of the United States, openly published by PJM Interconnection LLC (PJM) — a regional network organization in the United States. This time series is non-stationary and is characterized by the presence of both a trend and seasonality. Prediction of the next values of the time series based on previous values and the constructed neural network model showed high approximation accuracy, which proves the effectiveness of the proposed method.

The article presents the results of a numerical study of the flow structure in a two-dimensional flat diffuser. A feature of diffusers is that they have a complex anisotropic turbulent flow, which occurs due to recirculation flows. The turbulent RANS models, which are based on the Boussinesq hypothesis, are not able to describe the flow in diffusers with sufficient accuracy. Because the Boussinesq hypothesis is based on isotropic turbulence. Therefore, to calculate anisotropic turbulent flows, models are used that do not use this hypothesis. One of such directions in turbulence modeling is the methods of Reynolds stresses. These methods are complex and require rather large computational resources. In this work, a relatively recently developed two-fluid turbulence model was used to study the flow in a flat diffuser. This model is developed on the basis of a two-fluid approach to the problem of turbulence. In contrast to the Reynolds approach, the two-fluid approach allows one to obtain a closed system of turbulence equations using the dynamics of two fluids. Consequently, if empirical equations are used in RANS models for closure, then in the two-fluid model the equations used are exact equations of dynamics. One of the main advantages of the two-fluid model is that it is capable of describing complex anisotropic turbulent flows. In this work, the obtained numerical results for the profiles of the longitudinal velocity, turbulent stresses in various sections of the channel, as well as the friction coefficient are compared with the known experimental data. To demonstrate the advantages of the used turbulence model, the numerical results of the Reynolds stress method EARSM are also presented. For the numerical implementation of the systems of equations of the two-fluid model, a non-stationary system of equations was used, the solution of which asymptotically approached the stationary solution. For this purpose, a finite-difference scheme was used, where the viscosity terms were approximated by the central difference implicitly, and for the convective terms, an explicit scheme against the flow of the second order of accuracy was used. The results are obtained for the Reynolds number Re = 20 000. It is shown that the two-fluid model, despite the use of a uniform computational grid without thickening near the walls, is capable of giving a more accurate solution than the rather complex Reynolds stress method with a high resolution of computational grids.

The article presented the computational algorithm developed to study chemical processes in the internal flows of a multicomponent gas under the influence of laser radiation. The mathematical model is the gas dynamics’ equations with chemical reactions at low Mach numbers. It takes into account dissipative terms that describe the dynamics of a viscous heat-conducting medium with diffusion, chemical reactions and energy supply by laser radiation. This mathematical model is characterized by the presence of several very different time and spatial scales. The computational algorithm is based on a splitting scheme by physical processes. Each time integration step is divided into the following blocks: solving the equations of chemical kinetics, solving the equation for the radiation intensity, solving the convection-diffusion equations, calculating the dynamic component of pressure and calculating the correction of the velocity vector. The solution of a stiff system of chemical kinetics equations is carried out using a specialized explicit second-order accuracy scheme or a plug-in RADAU5 module. Numerical Rusanov flows and a WENO scheme of an increased order of approximation are used to find convective terms in the equations. The code based on the obtained algorithm has been developed using MPI parallel computing technology. The developed code is used to calculate the pyrolysis of ethane with radical reactions. The superequilibrium concentrations’ formation of radicals in the reactor volume is studied in detail. Numerical simulation of the reaction gas flow in a flat tube with laser radiation supply is carried out, which is in demand for the interpretation of experimental results. It is shown that laser radiation significantly increases the conversion of ethane and yields of target products at short lengths closer to the entrance to the reaction zone. Reducing the effective length of the reaction zone allows us to offer new solutions in the design of ethane conversion reactors into valuable hydrocarbons. The developed algorithm and program will find their application in the creation of new technologies of laser thermochemistry.