   |
  |
||
|
|||
 |
|||
“Systems and Means of Informatics” scientific journalVolume 31, Issue 1, 2021Content Abstract and Keywords About Authors CONCORDANT MODELS FOR LATENT SPACE PROJECTIONS IN FORECASTING
Abstract: The paper examines the problem of predicting a complex structured target variable. Complexity refers to the presence of dependencies, whether linear or nonlinear. The source data are assumed to be heterogeneous. This means that the spaces of the independent and target variables are of different nature. It is proposed to build a predictive model that takes into account the dependence in the input space of the independent variable as well as in the space of the target variable. It is proposed to make a model agreement procedure in a lowdimensional latent space. The projection to the latent space method is used as the basic algorithm. The paper compares the linear and proposed nonlinear models.
Keywords: prediction; partial least squares; model concordance; nonlinear projection to latent space
ON THE BOUNDS OF THE RATE OF CONVERGENCE FOR Mt/Mt/1 MODEL WITH TWO DIFFERENT TYPES OF REQUESTS
Abstract: The author deals with a nonstationary queuing model Mt/Mt/1 with one server and two different types of requests. For this model, the author obtains a one-dimensional birth and death process that describes the number of requirements in the original system. By applying the standard method of the logarithmic norm of the operator of a linear function, corresponding estimates for the rate of convergence and ergodicity are obtained. A numerical example with exact given values of intensities showing the application of the studied approach is constructed and corresponding graphic illustrations are provided. The author uses the general algorithm to build graphs, it is associated with solving the Cauchy problem for the forward Kolmogorov system on the corresponding interval which has already been used by the authors in previous papers. Keywords: queuing systems; nonstationary queuing model; one-dimensional birth-death process; rate of convergence; ergodicity bounds; logarithmic norm; Mt/Mt/1 queue
ON APPROXIMATION WITH TRUNCATIONS FOR THE NONSTATIONARY QUEUING MODEL
Abstract: The author deals with a nonstationary queuing model Mt/Mt/1 with one server. It is assumed here that the customers arrive with the intensity Л^) but are served in pairs (that is, in this case, yu(t) is the service rate of a group of two customers). For the considered model, the limiting characteristics are constructed using the method of truncating the state space of the system. A numerical example with exact given values of intensities showing the application of the studied approach is constructed and corresponding graphic illustrations are provided. The author uses the general algorithm to build graphs, it is associated with solving the Cauchy problem for the forward Kolmogorov system on the corresponding interval which has already been used by the author in previous papers. Keywords: queuing systems; Mt/Mt/1 queue; nonstationary queuing model; approximation; limiting characteristics; rate of convergence; truncation of the state space
ANALYTICAL MODELING AND FILTERING FOR INTEGRODIFFERENTIAL SYSTEMS WITH UNSOLVED DERIVATIVES
Abstract: For nonlinear integrodifferential stochastic systems (IDStS) with unsolved derivatives reducible to differential stochastic systems (StS) by means of singular kernels, the following methods and algorithms are proposed: analytical modeling of normal (Gaussian) stochastic processes and analytical synthesis of normal suboptimal filters for information processing in IDStS. Both Gaussian and non-Gaussian StS white noises are considered. Quality estimation methods based on the sensitivity theory are suggested. An example with discontinuous nonlinearity is considered in details. Directions for future investigations are given. Keywords: integrodifferential stochastic system (IDStS); method of analytical modeling (MAM); method of normal approximation (MNA); method of statistical linearization (MSL); normal suboptimal filter (NSOF); stochastic system (StS); stochastic systems with unsolved derivatives
RESEARCH AND DEVELOPMENT STRATEGY IN THE FIELD OF ARTIFICIAL INTELLIGENCE I: BASIC CONCEPTS AND BRIEF CHRONOLOGY
Abstract: The paper begins a series of works presenting the results of the study of the impact of public administration on the effectiveness in the field of artificial intelligence research and development (AI R&D), which has become a strategically important industry in any technologically developed country.
Keywords: artificial intelligence; related concepts and technologies of artificial intelligence; time chronology
SUPPORT FOR SOLVING DIAGNOSTIC TYPE PROBLEMS
Abstract: Some significant features of mathematical methods of data analysis and decision support in diagnostic-type problems are discussed. The most significant characteristic features are considered allowing to distinguish the tasks of the discussed type into a special class. This class requires the simultaneous development of solutions to a number of interrelated problems which, without taking into account such relationships, are practically useless. Using the experience with diagnostic-type tasks, recommendations are made on the areas of development of effective approaches and methods of data mining for solving such applications. Keywords: artificial intelligence; intelligent data analysis; mathematical methods; diagnostics
NEURAL NETWORK APPROACH FOR INFORMATION AND ANALYTICAL SUPPORT OF CONTROL AND PROTECTION OF AQUATIC BIOLOGICAL RESOURCES
Abstract: The article deals with the use of artificial neural networks (ANN) to solve some of the information and analytical support problems of goal-setting and situational management processes in the control and protection of aquatic biological resources (ABR) system. The analysis of this subject area allows one to identify a number of high-tech applied tasks of information and analytical support for the control and protection of ABR, primarily related to goal setting, calculation of forces and means, as well as their situational management. The classification of goal-setting, planning, and situational management tasks in this area is carried out. The structure and composition of the initial input data and outputs of two types of ANN - classification and forecasting - are justified. Issues of training and testing of neural systems are discussed. Keywords: artificial neural networks; aquatic biological resources; information and analytical support
ASSESSMENT OF THE EFFECT OF PROCESSES AND THREADS AFFINITY IN IBM POWER COMPUTING SYSTEMS ON THE PARALLEL APPLICATIONS PERFORMANCE
Abstract: The article is devoted to the study of the performance of computing systems based on modern IBM POWER processors when running parallel applications. The effect of different methods of distributing computational processes and threads among central processors' cores on the efficiency of executing programs developed using MPI (Message Passing Interface) and OpenMP (Open Multi-Processing) technologies is studied. The results obtained be useful for evaluating the efficiency of algorithms for the distribution of computing resources and the organization of the computing process in distributed heterogeneous computing systems. Keywords: heterogeneous computing; IBM POWER8; IBM POWER9; simultaneous multithreading; threads affinity; processes affinity; OpenMP;
MPI; NAS Parallel Benchmark
EVOLUTION OF NETWORK PROCESSORS
Abstract: Network processors (NP) have passed a long evolution way from general-purpose computers equipped with network interfacing cards to highly integrated on-chip systems comprising dozens of programmable processors, hardware accelerators, and network interfaces. The high performance was reached in the NPs on two differing directions: old microelectronic firms scaled up the number of processing cores with a conventional architecture, while small young companies invented specialized approaches. The second direction provided better relative characteristics, but the small firms were not competitive due to complexities of programming in their nonconventional architectures and paucity of resources to create comprehensive NP software. Ultimately, almost all of them left recently the NP market, having been absorbed, in one way or another, by some larger companies. Herewith, the latter utilized the acquired technologies mostly not for expansion into the NP market so much as for applying the multicore experience to conquer the high-performance server processors market which rapidly grows in the nowadays' era of cloud technologies. Thus, the NPs are actually ceasing to exist as separate products and turning into an intrafirm tool for intellectualization of a restricted circle of some specific network devices. Keywords: network processor; integrated network processor; processor core architecture; high-performance multicore server
METHODS FOR COMPARING COMPETING HYPOTHESES IN HYPOTHESIS-ORIENTED SYSTEMS
Abstract: With the advent of a new class of virtual experiments management systems, the use of hypotheses and models in an explicit form becomes more and more widespread. Such systems apply both hypotheses generated from the data and theoretical hypotheses. It becomes critically important to compare several competing hypotheses of different origin with each other. The paper considers various approaches to comparing competing hypotheses and computational models implementing them. The considered approaches are implemented as a software component that is a part of a virtual experiment management system. The component is applied for problem solving in neurophysiology. Keywords: virtual experiments management systems; competing hypotheses; comparison of hypotheses
EMPLOYING DEEP LEARNING NEURAL NETWORKS IN MATHEMATICAL BASIS OF DIGITAL TWINS OF ELECTRICAL POWER SYSTEMS
Abstract: Development problems for digital twins of contemporary active power distribution systems are considered. Approaches to employing deep learning neural networks in digital twin-based intelligent control of these systems are highlighted. A brief review of relevant neural network architectures is outlined. Examples of neural network tools for solving a number of key intelligent control problems are presented, including load forecasting, electricity price forecasting, economic dispatch, power machine health assessment and prediction, and faults and disasters diagnosis. Recommendations are provided regarding alternative deployment modes of the presented neural network tools, such as inclusion into a digital twin basic mathematical software, or supply as auxiliary applications for certain categories of users. Keywords: digital twin; electrical distribution system; deep learning neural network; forecasting; fault diagnosis
THE MODEL OF COMMUNITY OF CONCRETE HISTORICAL INVESTIGATION SUPPORT TECHNOLOGY USERS
Abstract: The article continues a series of works devoted to description and analysis of distributed technology of concrete historical investigation support based on the principles of crowdsourcing. This article is devoted to description and substantiation of the approach to modeling the community of technology users and the processes of spreading distorted information among its members to study the resistance of technology of concrete historical investigation support to attempts of history distortion, which is an urgent task in modern society. The proposed approach is to create a model of community structure on the basis of the Bollobas-Riordan directed graph. The principles of information spreading between graph nodes are based on factors and effects that occur in real social networks and are determined by the characteristics of their members, the nature of their interaction, and the properties of the social network. The adequacy of the model was verified by comparing the character of distorted information spread in the mode of resistance to information attacks lack with the logistic curve that shows the process of innovation diffusion. Keywords: virtual community; model; technology; distortion of history; concrete historical investigation
AN APPROACH TO IMPROVING THE CONCEPTUAL SCHEMES OF GEODATA BY MEANS OF MODELS FOR SPATIAL AND LOGICAL LINKING OF GEOOBJECTS
Abstract: In the context of research on improving the conceptual schemes for the representation of topographic information to develop the processes of geoanalytics in promising geographical information systems, one of the problems that occur in traditional conceptual schemes is considered here. An extended typology of types of spatial localization of geoobjects, including composite geoobjects, is presented. As a solution, it is proposed to introduce special data objects into the structure of the geodata databases which provide an explicit spatial-logical link between the components of composite geoobjects and functional infrastructures. Keywords: conceptual schemes of geodata databases; types of spatial localization of geodata; spatial-logical linking of geoobjects
SIR-MODEL AS A TOOL TO STUDY DESTRUCTIVE PROCESSES IN NEW KNOWLEDGE ACQUISITION
Abstract: The article describes an approach to the analysis of mastering new knowledge with the use of mathematical modeling of a learning situation. The proposed model is based on W. Kermack and A. McKendrick's SIR (susceptible- infected-recovered) model which was originally used to predict the spread of an epidemic to large closed populations in order to prevent the disastrous consequences of global infection. The modified SIR-model allows researchers to investigate a number of regularities that a cognitive process carried out in a closed small student community has. The model also aims at identifying the ways of behavior of such a dynamic system when working with excessive information can lead to cognitive overload and errors in solving learning tasks. The approach is implemented as a numerical and graphical experiment on the phase plane, which makes it possible to compose a holistic picture of the phenomenon and to analyze the conditions under which the system moves towards the state of stable equilibrium. The latter is equal to overcoming cognitive overwork by the doers. Keywords: cognitive process; new knowledge; cognitive overwork; closed student community; Kermack-McKendrick model; SIR-model; phase plane/portrait; system stability/instability
THE MODEL OF NORMALIZED ECONOMICS AND RELEVANT TECHNOLOGIES OF DIGITALIZATION
Abstract: The review presents key updates of the model of normalized economics and relevant technologies of digitalization. The updated model is presented by description of the normalized economic mechanism (NEM) which should be implemented on the basis of online services operating in the environment of digital twins. Normalized economic mechanism includes the situationally managed systems of resource support, production of real goods, etc. The NEM banking system includes personal electronic banks of individuals, corporate electronic banks, provider banks, and the regulation bank under which all other banks operate. The normalized money is intended to represent the value of goods and property statuses of the economic activity participants, to pay for goods, invest, and accumulate wealth. Normalized money serves as a universal electronic means of quantitative documentation of property relations to be certified by the state through the online services of the regulation bank. An economic management is to be based on the systems of mandatory and orienting requirements to the results of solving problems inside the NEM complexes. Keywords: model of normalized economics; normalized economic mechanism; personal electronic banks; corporate electronic banks; designated payments technology; online cost planning service
|
 | |
 Phone of the Center: +7 (499) 135-62-60E-mail of the Center: ipiran@ipiran.ru | 
RUS
|