ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 2. Vol. 21. 2015

To the contents

V. P. Kulagin, Deputy Director in Innovations, Professor, e-mail: vkulagin@hse.ru, National Research University Higher School of Economics (HSE)

Tensor Methods of Research Structures of Petri Nets

The article describes the tensor approach to the study of complex systems in terms of Petri nets. Introduced the concept of different systems, which represent the original SP-structure and its derivatives in different coordinate systems. It is shown that using tensor methods, greatly simplifies the procedure of construction of possible structures of the studied complex systems in the coordinate system of the primitive system, and makes routine transformation of new structures of complex systems in the original coordinate system. The described approach provides new opportunities to build methods for synthesis of structures of complex systems.
Keywords: tensor calculus, Petri nets, patterns of complex systems, coordinate system

P. 83—94

A. A. Sirota, Professor, A. V. Tsurikov, Postgraduate Student, e-mail: andrew.tsurikov@gmail.com, Voronezh State University

Creating Content-Dependent Digital Watermarks Using their Structural Patterns: Models and Algorithms of Text Fragments Classification

The paper introduces approaches to creating content-dependent digital watermarks for the text data containers. It proposes ways of encoding text for further implementation of the matching procedure between the encoded text fragments and binary encoded digital watermark elements. The paper shows that this matching procedure is fully equivalent to the classification of the high-dimensional data. It also outlines approaches to creating high-dimensional data classifiers using various techniques for developing content-dependent digital watermarks, and compares them with other classification algorithms (neural networks, support vector machines and potential functions). Along with common algorithms of machine learning, it considers the ways of making them more efficient by limiting the number of the elements that take part in the classification. It also examines the dependency between the data classification errors and the space dimension size for different classifiers using simulated and real text data.
Keywords: content-dependent digital watermark, data classification, radial-basis function, support vector machine, potential functions method, neural networks

P. 95—103

B. G. Kukharenko, Leading research scientist, Blagonravov Institute of Engineering Science of RAS, M. O. Solntseva, Post-graduate student, Moscow Institute of Physics and Technology (SU)

Analysis of Multi Dimensional Trajectory Clustering by Models of Linear Dynamical System

For clustering multidimensional trajectories polynomial regression method with parameter leaning by the Expectation-Maximization algorithm is in use. The method based on polynomial regression is characterized by the joint clustering and continuous alignment of curve sets in time and space. Nevertheless, a number of defined clusters can't be large enough. Thus, for a set of sufficiently heterogeneous trajectories, the defined clusters are heterogeneous also. For the inhomogeneous cluster, its polynomial regression is a too strong abstraction. To demonstrate the cluster heterogeneity (and, thus, non full clustering trajectories) a method of dimension reducing is in need. So, linear dynamical system models are applied to object multi dimensional trajectory clustering by polynomial regression method. An advantage of linear dynamical systems is reducing clustering result dimension. For trajectory coordinate projections in a cluster most informative component (polynomial regression) is extracted and the cluster fine structure is appeared. An efficiency of linear dynamical systems is demonstrated by example of clustering results of airplane flight tracks in an airport space.
Keywords: data mining, multi-dimensional trajectories, clustering, polynomial regression, Kalman filter, Rauch smoother, linear dynamical systems, Expectation-Maximization algorithm

P. 104—109

R. V. Kazakov, Lecturer, Bryansk State Technical University

The use Differential Evolution to Obtain Pareto-Set by the Multi-Objective Genetic Algorithms

The new manner to improve the efficiency of multi-objective genetic algorithms is considered. It is based on the principles of differential evolution in the process of creating new individuals. The use of general schemes of differential evolution DE/rand/1/bin etc. for multi-objective optimization are analyzed. It is drawn a conclusion that the general schemes mainly use the information about location of solutions in a decision space that is often insufficient for effective exploration of the objective space. Therefore the original scheme of differential evolution DE/rand/1X/bin is suggested. By this scheme in recombination process is taken an account of searching path into two spaces — decisions and objectives. This process is based on the rule that the selection of second parent and other (two or more) individuals is made randomly. Hence to increase the probability of creating individuals that dominate their parents various combinations of individuals selected for recombination may be examined. For example in general scheme DE/rand/1/bin it means that each of three randomly selected individuals can be selected as a second parent. If no one of various combinations of individuals in recombination process does not allow to get individual-offspring being better than their parents, the parent-individual one is selected for next generation. It was found that such scheme should allow to realize higher replacement of the population dominated individuals. Also it was found that the number of trials for offspring creation self-adapted and changed according to array of current population. The scheme DE/rand/1X/bin should be integrated into other multi-objective genetic algorithms to handle the obtained nondominated solutions. The experiments for evaluation scheme DE/rand/1X/bin were conducted on the benchmark problems DTLZ. The results was found that the use of such a scheme allowed to improve the results (on a some indicators) of algorithms SPEA2, NSGA-II and their modifications with a general scheme of differential evolution.
Keywords: multi-objective optimization, Pareto's principles, Pareto front, multi-objective genetic algorithms, differential evolution

P. 109—116

A. N. Vdovin, Senior researcher, "TINRO-CENTER", Vladivostok, e-mail: vdovin@tinro-center.ru, A. N. Chetyrbotsky, Senior researcher, Far-East Geology Institute RAS, Vladivostok, chetyrbotsky@yandex.ru, V. A. Chetyrbotsky, Student, Moscow University, Moskva, ve14232@gmail.com

Mathematical of the Dynamics Fish Growth (for example, southern Atka mackerel Pleurogrammus azonus)

As part of the system of "resource—consumer" is developed the model of dynamics of fish growth. To construct it, we used: the maximum length of the fish, the stages of their life cycle and the impact of seasonal factors. The parametric identification of the model is performed. According to her is carried out the assessment of the statistical properties of the model parameters, is established the adequacy between the empirical distribution and its model image. On the based of computational experiments the assessment the assessment of duration of the life cycle period.
Keywords: growth dynamics, the logistic equation, parameter identification, the task of finding the minimum value model

P. 116—120

V. B. Kovalenko¹, Senior Researcher, e-mail: vereten@hotbox, A. I. Dordopulo², Head of Laboratory, e-mail: scorpio@mvs.tsure.ru, V. A. Gudkov², Senior Researcher, e-mail: gudkov@mvs.tsure.ru, A. A. Gulenok², Senior Researcher, e-mail: andrei-gulenok@mail.ru, L. M. Slasten², Senior Researcher, e-mail: emslasten@yandex.ru
¹Southern Scientific Centre of the Russian Academy of Sciences
²Academical A. V. Kalyaev Scientific Research Institute of Miltiprocessor Computer at Southern Federal University

Use of Soft-Architectures for Digital Signal Processing Tasks Solved on Reconfigurable Computer Systems

The paper covers the macroobject approach to reconfigurable computer system (RCS) programming, which reduces the programming time of tasks owing to reduction of the translation time of parallel programs. The description of the main advantages and shortcomings of the existing RCS programming levels are given. We have analysed the notion of a soft-architecture, a macroobject, a node and objects of reconfigurable computer systems. We have given the structure of system software owing to which the suggested approach to RCS programming is provided. In addition we have described functions of the elements of the software suit and their interconnection during development of soft-architectures, and use of soft-architectures for implementation of application tasks. The development of the soft-architecture for digital signal processing tasks is analysed, and examples of SADL-descriptions of soft-architecture elements are given. The example of the soft-architecture used for implementation of digital signal processing tasks is given. Besides, we have given the description of one cadr of the parallel application, written in the high-level programming language COLAMO. It implements the FFT algorithm on the base of the DSP soft-architecture.
Keywords: multiprocessor systems, supercomputers, reconfigurable computer systems, architectures of computer systems, parallel programming, programming languages

P. 121—127

À. V. Poskonin, Postgraduate, e-mail: aposk@yandex.ru, Lomonosov Moscow State University

Integrating SQL-Based DBMSs with NoSQL Datastores at the Object-Mapping Layer

This paper will discuss and evaluate an approach to building an object-mapping layer that provides integration of multiple different data management systems, including SQL-based DBMSs and popular NoSQL systems. A prototype object-mapping framework was developed to illustrate basic concepts of such integration including unified entity-based query language, transparent object mapping and a high-level interface, that still allows low-level optimizations and delivers sufficient performance for most use cases. Unified query language makes it possible to translate a single query to different underlying datastores without a need to modify the query while objectmapping layer manages object construction and lifecycle, providing simple interface for persisting, modifying and deletion of objects. In conclusion, performance of the implemented framework is measured and compared with popular ORM solution.
Keywords: SQL, NoSQL, polyglot persistence, service-oriented architecture, object-relational mapping, object-document mapping, integration

P. 128—132

Ad. M. Dimov, Professor, O. N. Maslov, Head of Chair, e-mail: maslov@psati.ru, Yu. V. Troshin , Associate Professor Povolzhskiy State University of Telecommunications and Informatics

Statistical Simulation Modeling Process Software Choice

The article considers the problem of statistical simulation modeling (SSM) software choice for complex organizational-technical systems. The examples of provided SSM-models applying universal language Delphi and simulation environment AnyLogic are presented. The SSM business process model of paying customers service of Telecommunications Company is the first example of a programming language Delphi. The second example — is SSM business process model of resuscitating (repairing) of oil wells. The "Order Fulfillment" business process of trading company modeled in the AnyLogic. Presented SSM models are destined for practical application by managers of telecommunication, oil-producing and trading business.
Keywords: statistical simulation modeling method, software, universal language Delphi, simulation environment AnyLogic

P. 132—139

S. V. Dvornikov¹, Professor, S. S. Manaenko¹, Associate Professor, S. V. Dvornikov², Student, A. A. Pogorelov¹, Head of the Department
¹Military Communications Academy, St. Petersburg
²St. Petersburg State Polytechnical University e-mail: practicdsv@yandex.ru

Synthesis PSK Wavelet-Signal

On improving immunity has always been given priority in the development of radio communication. Among the well-known modulation formats is the most error-correcting binary phase shift keying is used extensively to radio HF VHF bands. Advances in the multiresolution analysis suggest that the synthesis signal on the basis of wavelets allow to obtain modulation formats with high noise immunity properties. It can be argued that the wavelet signals provide additional structural secrecy in relation to systems analysis using other functional bases of processing radio signals.
In this regard, the article discusses the synthesis of phase-shift keyed signals based on wavelet and study their immunity in an additive white Gaussian noise.
The very same synthesis procedure is as follows. Forming a carrier wave of a certain frequency, which in accordance with a predetermined manipulation speed is changed according to the value of the phase information sequence of logical ones and zeros. As the chip is proposed to use fragments of first-order Gaussian wavelet that represents the first derivative of the Gaussian function.
Localization properties of the wavelet in time and bilateral temporal structure allow it to form the basis of fluctuations, which can also be regarded as a phase-shift keyed signals.
To estimate the noise immunity of the proposed PSK-2 wavelet design was defined channel with AWGN. As the initial capacity of the fragment was determined based on the sine wave signal generated by the quadrature. Amplitude sine wave was timebase . To provide similar facilities for fragment-based wavelet, its amplitude was increased to a level of 2,84. High immunity to ensuring wavelet signals only when the receiving end of the correlation processing is performed in the wavelet base functions.
Synthesis of signals based on the fragments of wavelets allows to obtain the modulated vibration immunity which exceeds 2 dB potentially possible indicators for PSK-2 signals. If considered as modulating fragments of first-order Gaussian wavelets, the win in order to provide additional structural secrecy of the order of 1 dB.
Further studies are seen in the development of effective methods of wavelet signal demodulation.
Keywords: synthesis signals, parametric stealth, Gauss wavelet first order demodulation of PSK signals

P. 140—143

A. I. Galushkin, Prof., Deputy Head of Chair, Moscow Institute of Physics and Technology, Dolgoprudny, neurocomputer@yandex.ru

Memristor in High-Performance Computing Development

The author of this article is not a physicist or a mathematician. He is an engineer and has 50 years of experience in neural network technologies and their application. This is a rather narrow class of over high performance computing. This article presents the author's opinion about the prospects of this trend of super computing with the memristor invention.
Keywords: memristor, supercomputer, multilayer, neural network, analog computation

P. 146—156

V. V. Fedosov, Associate Professor, e-mail: vlr.fdsv@gmail.com, A. V. Fedosova, Universidad Nacional de Colombia, e-mail: afedosova@unal.edu.co

The Use of Neural Functions for Graphical Evaluation Contamination Emissions Group Sources

Under the control of pollution of the territory, region or environment as a result of emissions from industrial sources, distributed monitoring, forecasts or estimates. Monitoring is necessary to spend, but few informative for understanding the relationship and management of complex systems. Projections should be based on a representative database, but the results are only probabilistic nature. Calculations pollution as complex systems that take into account a large number of variables that are to manage the impact on one or more of the evaluation criteria.
In the simulation of industrial ecology is a key functional description of the emission sources. His concept of the mathematical apparatus, the introduction of complications (simplications) determined by the objectives and tasks of the qualifications of the investigator. However, authoring mathematics is often subjective, is quite laborious and limited variability describe clouds emissions. Need a tool cloud formation emissions without mathematical formulas, but lets you embed cloud monitoring data and forecast pollution.
Such opportunities opened by the use of neural networks. This results in: the abolition of the author of mathematics; considerable simplification of the input database model; work directly with clouds emissions; automatically obtain the functional description emissions.
In this paper we propose a simple and fast way of learning neural function according cartoon clouds contour plots emissions. Shows examples of the formation of clouds emissions of different configurations. Neural function effectively graphics output selective or general pollution of the territory and are suitable for further calculation or optimization.
Keywords: sources of industrial emissions, cloud emission, pollution area, contour plots, neural Networks, filtering emissions

P. 156—160

To the contents