2017 № 1(34)
Contents

Zhusupbaev A., Toktoshov G.Y. ONE PROBLEM OF THE OPTIMIZATION RESOURCE DISTRIBUTION IN HIERARCHICAL NETWORKS

Rodionov A.S. CUMULATIVE ESTIMATED VALUES OF STRUCTURALNETWORK'SRELIABILITY INDICES AND THEIR USAGE

Biyashev R.G. , Kalimoldayev M. N., Rog O.A. TYPED ATTRIBUTEBASED ACCESS CONTROL SEMANTICS MODELING

Khairetdinov M.S., Poller B.V. , Britvin A.V., Sedukhina G.F. ACOUSTICOPTICAL SYSTEM OF ULTRALOW FREQUENCIES

Kholmonov S.M. OPTIMIZATION OF DATA PROCESSING ON THE BASIS OF MODIFIED TRAINING OF NEUTRAL NETWORK AND SEGMENTATION OF CASUAL TIME PROCESSES

Moiseenko V.V., Rodionov A.S. AGE DEPENDENCY OF COAUTHORSHIP IN SCIENTIFIC RESEARCHES IN ACADEMIA

Malyshkin V.E., Perepelkin V.A., Schukin G.A. DISTRIBUTED ALGORITHM FOR DATA ALLOCATION IN LUNA FRAGMENTED PROGRAMMING SYSTEM
ONE PROBLEM OF THE OPTIMIZATION RESOURCE DISTRIBUTION IN HIERARCHICAL NETWORKS
UDC 519.8:624.9
The paper deals with one of the applied problems in the design and modernization of the distribution points of the various services (eg, gas, oil, and water)  the problem of the optimal distribution of production and distribution of resources between consumers and suppliers, taking into account a variety of restrictive conditions. In this regard, the topic under consideration, relating to the design and development of the mathematical apparatus to describe, analyze, and optimize network engineering maintenance vehicle type is relevant.
In this paper, engineering support network is seen as a hierarchical system of the vehicle type, twostage, respectively solved the problem of accommodation with a nonlinear objective function. It is assumed that resources (oil, gas, water, etc.) are transferred in this system from producers to consumers through processing by points connections (transport channels). Such systems are characterized by the presence of socalled fixed costs to be made regardless of the target volume roduction and processing (gas, oil, etc.). The account in the models of these and similar factors leads to the appearance of their objective functions that do not possess the property of continuity. Therefore, the problem of resource allocation in
hierarchical networks, is seen as a twostage problem of placing items resource production and distribution with discontinuous objective function. This class of problems, which bears the character of multi considerably complicates the search for an extremum of the objective function, and can not be solved in the general case. Thus, there is a need to develop a new method of solving the problem, when the functions that determine the cost of production, transportation and processing of products are nonlinear.
In this regard, the aim of this work is to develop new mathematical models and methods for analysis and optimization of engineering support transporttype networks with a hierarchical structure. The goal is achieved as a result of research and development of the mathematical apparatus and optimization techniques for the design and optimization of engineering support such as transport networks, taking into account the various restrictive conditions of feasibility, natural, nvironmental, situational and character, as well as legal and budgetary constraints of the region.
Note that the distribution of production and resource allocation problems have a wide range of applications that arise when designing of engineering networks for different purposes, in problems of logistics facilities and power systems, planning and reconstruction transporttype networks, routing service networks and other areas.
The paper studied and proved a sufficient condition for the applicability of the method of successive twostep calculation for the problem of locating the production and processing of nonlinear functions of transport, operating costs and xpenses for processing at various restrictive conditions. A combinatorial method for optimizing the placement based on the method of consecutive calculations, which allows to calculate the best option to deploy elements of communications and network engineering plan (volume) of the transported goods from sources to consumers by the criterion of minimum total cost of production of the raw material, its processing and transportation. The proposed method makes it possible, though, and by the partial sorting, finding the global extremum multiextremal problem with any degree of accuracy.
The proposed methods and models allow to carry out a reliable analysis and optimization of engineering support networks and can be used in the design organizations involved in the design, construction and operation of networks for different purposes.
Key words: networks location problem, the method of successive calculations, placement plan, the optimal output of production.
2. Melkumov V. N., Kuznetsov I. S., Kuznetsov R. N. Opredelenie optimalnogo marshruta trassy gazoprovoda na osnove kart stoimosti vliyayushchikh faktorov [Determination of the optimal route of the gas pipeline on the basis of the factors affecting the value of cards] // Nauch. vestn. VGASU. Stroitelstvo i arkhitektura. 2009. N 1 (13). P. 21–27.
3. Kolokolov A. A., Kosarev N. A. Issledovanie otsecheniy Bendersa dlya zadachi o pmediane [Baderses clipping research for the problem of pmedian]// International conference Optimization problems and economic applications. Omsk. 2003. P. 96.
4. Kochetov JU. A. Zadacha ob (r,p)cenroide [The problem of (r,p)centroid] // International conference Optimization problems and economic applications. Omsk. 2009. P. 68.
5. Sarychev D. S. Sovremennye informatsionnye sistemy dlya inzhenernykh setey [Modern information systems for engineering networks] // Vestnik TGU. 2003. N 280. P. 358–361.
6. Skvortsov A. V. Razrabotka geoinformatsionnykh system na fakultete iformatiki I v OOO Indorsoft [Development geographic information and engineering systems at the faculty of Informatics and OOO Indorsoft] // Vestnik TGU. 2003. N 280. P. 346–349.
7. Boikov V. N., Shumilov B. M. Splainy v trassirovanii avtomobilnykh dorog [The splines in the routing of roads]. Tomsk. 2001.
8. Lempert A. A. Matematicheskaya model I programmnaya sistema dlya razmesheniya zadach logisticheskikh obektov [Mathematical model and software system for the solution of location problems of logistics objects] / A. A. Lempert, A. L. Kazakov., D. S. Bukharov // Management of big systems. 2013. N 41. P. 270–284.
9. Kosyakov S. V. Myagkie vychisleniya v postrornii kart zonirovaniya territorij po parametram system energosnabzheniya [Soft computing in mapping zoning by the parameters of power supply systems] // S. V. Kosyakov, S. S. Novoseltseva, A. M. Sadykov // International scientifictechnical conference The state and prospects of development of electrotechnology. Ivanovo, 2013. P. 338–341.
10. Prilutskiy M. X., Afraiymovich L. G. Raspredelenie resursov v ierarkhicheskikh sistemakh transportnogo tipa: uchebnoe posobie [Distribution of resources in hierarchical systems of transport type: textbook] / M. X. Prilutskiy, L. G. Afraiymovich. Nizhny Novgorod. 2007.
11. Erzin A. I. Zadachi marshrutizatsii: uchebnoe posobie [The routing problem: textbook] / A. I. Erzin, JU. A. Kochetov. Novosibirsk, 2014.
12. Lange E. G., Zhusupbaev A. Kombinatorniy metod resheniya zadachi razmesheniya [Combinatorial method for solving location problem]. Frunze, 1991.
13. Cherenin V. P. Reshenie nekotorykh kombinatornikh zadach optimalnogo planirovaniya metodom posledovatrlnikh raschetov [The solution of some combinatorial problems optimal planning by methods of successive calculations]. Moscow, 1962.
14. Zhusupbaev A., Sharshembieva F. K. Reshenie dvukhetapnoy zadachi razmesheniya s nelineynoy tselevoy funktsiey [The solution of the twostage location problem with nonlinear objective function] // Vestnik KNU. 2013. Vypusk 1. P. 21–29.
15. Asankulova M., Zhusupbaev A. Optimizatsiya dobichy I raspredeleniya sirya mezhdu potrebitelyami v zavisimosti ot perioda[Resources extraction and distribution optimization between consumers depending on the period] / Modern problems of science and education. 2016. N 4. P. 7–12.
Bibliographic reference: Zhusupbaev A., Toktoshov G.Y. One problem of the optimization resource distribution in hierarchical networks //journal “Problems of informatics”. 2017, № 1. P. 214.
Institute of Computational Mathematics and Mathematical Geophysics of SB RAS, 630090, Novosibirsk, Russian Federation
CUMULATIVE ESTIMATED VALUES OF STRUCTURALNETWORK'SRELIABILITY INDICES AND THEIR USAGE
UDC 519.17
Tasks of obtaining and usage of exact cumulative expected values of some indices of a network's structural reliability, such as kterminal probabilistic connectivity (kTPC), average pairwise connectivity (APC) and expected size of a subnetwork that contains some special node (MENC) are considered in the paper. While first index is well known and explored, last ones are not so. Note that APC characterizes network from a point of view of uses (it corresponds to a probability of stablishing an arbitrary connection), and MENC is very important for stimating quality of a sensor or other monitoring network. Random nonoriented random graph with reliable nodes and unreliable edges that fails independently is used as a model of a network. At the same time the proposed approach can be used in the case of unreliable nodes also. The factoring method, exhaustive search, methods of reduction and decomposition are used as a base. Expected values are used for unambiguous decision making about network's reliability, designing evolutionary algorithms for structural optimization and obtaining approximate expected values that are more precious than cumulative ones obtained to the moment.
Exact cumulative expected values (lower  LB, and upper  UB) change (at least one of them) after obtaining value of an index used for some new realization of a network's structure in such a way, that after obtaining value of this index for a last possible realization both LB and UB become equal to an exact value of an index under consideration. The main idea is rather obvious one: when some values of an index with probabilities of corresponding realizations are obtained, we assume that all other realizations have possible minimal (for LB) or maximal (for UB) value of an index. This guarantees that exact value lays between LB and UB and that after last step LB, UB and exact value of and index are equal.
Key words: network reliability, reliability indecies, estimation, optimization.
2. Yano A. and Wadayama T. Probabilistic analysis on network reliability problem. CoRR. V. abs/1105.5903, 2011. [Online]. Available: http://arxiv.org/abs/1105.5903.
3. Tauro S. L., Palmer C. R., SigaN G., and Faloutsos M. A simple conceptual model for the internet topology // Proceedings of the Global Telecommunications Conference, 2001. GLOBECOM ’01, San Antonio, TX, USA, 25–29 November, 2001. IEEE, 2001. P. 1667–1671. [Online]. Available: http://dx.doi.org/10.1109/GLOCOM.2001.965863.
4. Shooman A. M. Algorithms for network reliability and connection availability analysis // Electro/95 International. Professional Program Proceedings. Jun. 1995. P. 309–333.
5. Sun F. and Shayman M. On the average pairwise connectivity of wireless multihop networks // Global Telecommunications Conference, 2005. GLOBECOM ’05. IEEE. V. 3. Nov. 2005. P. 5.
6. Rodionov A. S. and Rodionova O. K. Exact bounds for average pairwise network reliability // The 7th International Conference on Ubiquitous Information Management and Communication, ICUIMC ’13, Kota Kinabalu, Malaysia. Jan. 17–19, 2013. ACM, 2013. P. 45. [Online]. Available: http://doi.acm.org/10.1145/2448556.2448601.
7. Valiant L. G. The complexity of enumeration and reliability problems // SIAM Journal on Computing, 1979. V. 8. N 3. P. 410–421. [Online]. Available: http://dx.doi.org/10.1137/0208032.
8. Bodlaender H. L., Bodlaender H. L., Wolle T. and Wolle T. A note on the complexity of network reliability problems. // IEEE Trans. Inf. Theory, 2004. V. 47. P. 1971–1988.
9. Deuermeyer B. L. A new approach for network reliability analysis // IEEE Transactions on Reliability, Oct 1982. V. R31. N 4. P. 350–354.
10. Goyal N. K., Misra R. B., and Chaturvedi S. K. Snem: a new approach to evaluate terminal pair reliability of communication networks // Journal of Quality in Maintenance Engineering. 2005. V. 11. N 3. P. 239–253. [Online]. Available: http://dx.doi.org/10.1108/13552510510616450.
11. Xie M., Dai Y., and Poh K. Computing systems reliability — models and analysis. Kluwer, 2004.
12. Won J. M. and Karray F. Cumulative update of allterminal reliability for faster feasibility decision // IEEE Transactions on Reliability, Sept 2010. V. 59, N 3. P. 551–562.
13. Rodionova O. K., Rodionov A. S., and Choo H. Network probabilistic connectivity: Exact calculation with use of chains // Computational Science — ICCS 2004, 4th International Conference, Part I, ser. Lecture Notes in Computer Science, M. Bubak, G. D. van Albada, P. M. A. Sloot, and J. Dongarra, Eds. Springer Berlin Heidelberg, 2004. V. 3036. P. 565–568.
14. Rodionov A. and Rodionova O. Network probabilistic connectivity: Expectation of a number of disconnected pairs of nodes // High Performance Computing and Communications, ser. Lecture Notes in Computer Science, M. Gerndt and D. Kranzlmüller, Eds. Springer Berlin Heidelberg, 2006. V. 4208. P. 101–109. [Online]. Available: http://dx.doi.org/10.1007/11847366_11.
15. Rodionov A., Rodionova O., and Choo H. On the expected value of a number of disconnected pairs of nodes in unreliable network // Computational Science and Its Applications — ICCSA 2007. Ser. Lecture Notes in Computer Science, O. Gervasi and M. Gavrilova, Eds. Springer Berlin Heidelberg, 2007, V. 4707, P. 534–543. [Online]. Available: http://dx.doi.org/10.1007/9783540744849_46.
16. Alexey D. A. M., Rodionov S. and Rodionova O. K. Improvements in the efficiency of cumulative updating of allterminal network reliability // IEEE Trans. Reliability, 2012. V. 61. N 2. P. 460–465. [Online]. Available: http://dx.doi.org/10.1109/TR.2012.2196172.
17. Wigderson A. The complexity of graph connectivity // Proceedings of the 17th Mathematical Foundations of Computer Science conference, Havel and Koubek (eds.), Lecture Notes in Computer Science 629, SpringerVerlag, 1992. P. 112–132.
18. Migov D. A., Nechunaeva K., Nesterov S., and Rodionov A. Cumulative updating of network reliability with diameter constraint and network topology optimization // Computational Science and Its Applications — ICCSA 2016, ser. Lecture Notes in Computer Science, Gervasi O., Murgante B., Misra S., Rocha A., Torre C., Taniar D., Apduhan B., Stankova E., and Wang S., Eds. Springer Berlin Heidelberg, 2016. V. 9786. P. 141–152.
Bibliographic reference: Rodionov A.S. Cumulative estimated values of structuralnetwork'sreliability indices and their usage //journal “Problems of informatics”. 2017, № 1. P. 1524.
Biyashev R.G., Kalimoldayev M.N., Rog O.A.
Institute of Information and Computational Technologies, 050010, Almaty, Republic of Kazakhstan
TYPED ATTRIBUTEBASED ACCESS CONTROL SEMANTICS MODELING
UDC 004.94
In this paper we put a definition, formulated principles and built a formal model for the typed attributebased access control and its version, namely multicriteria typed attributebased access control. Hierarchical semantic modeling system has been developed allowing designing different models of access control. It consists of a metamodel, specific models obtained from it, and access matrices corresponding to that models. The proposed typed attributebased access control method is based upon the concept of actively developed now attributebased access control (ABAC). The essence of ABAC lies in the fact that the entities are assigned attributes which form access control criteria. Developing solutions for the subject to access an object is based on the calculation of attribute values. The positive quality of ABAC is its universality and flexibility. Hence, the ABAC can serve as a means to create different access control policies performing tasks of protection of information, which are characteristic to a specific organization. In particular it can represent such the most widely spread models as DAC, MAC, RBAC. On the other hand, the complexity of the implementation and lack of formal models pointed out by many researchers hamper the wide use of ABAC.
We propose a formal model for typed attributebased access control model and a corresponding model for multicriteria typed attributebased access control. The idea of this model is based upon a notion of a type of the access control attributes. The type defined on several levels as a mathematical object, constitutes a semantic modeling system, which is intended to assign the semantic values to the names of the model. First, it assigns values to the names of the metamodel, thus forming a certain access control model, such as DAC, MAC or RBAC. Then during the stage of the functioning of the access control system based upon the model obtained, the semantic modeling system assigns values to the attributes of the entities which belong to the access control matrix. Performing operations on the attribute values of the same type as well as on their structured subsets reduce the computational complexity, speeds up the calculation and facilitates the administration process. Systems containing several independent types are able to perform multicriteria access control.
Key words: access control task subject area, semantics, typed attributebased access control, attribute type, semantic modeling system, metamodel, access control model, access matrix.
2. Hu V.C., Ferraiolo D., Kuhn R., Schnitzer A., Sandlin K., Miller R., Scarfone K. NIST Special Publication 800–162. Guide to Attribute Based Access Control (ABAC). Definition and Considerations // NIST National Institute of Standards and Technology. [Electron. res.]. http://dx.doi.org/10.6028/NIST.SP.800162. Jan. 2014.
3. Jin X., Krishnan R., Sandhu R. A unified attributebased access control model covering DAC, MAC and RBAC // Proceedings of the 26th Annual IFIP WG 11.3 conference on Data and Applications Security and Privacy (DBSec’12). 2012. P. 41–45.
4. Bijon Kh., Krishnan R., Sandhu R. Constraints Specification in Attribute Based Access Control // IEEE/ASE Science Journal. 2013. V. 2. N 3. P. 131–144.
5. Biyashev R. G., Kalimoldaev M. N., Rog O. A. Polimorfnaya tipizaciya sushchnostej i zadacha konstruirovaniya mekhanizma mnogokriterial’nogo razgranicheniya dostupa // Izvestiya NAN RK. Seriya fizikomatematicheskaya. 2014. N 5. P. 33–41.
6. Biyashev R. G., Kalimoldaev M. N., Rog O. A. Konstruirovanie sistem mnogokriterial’nogo atributnogo razgranicheniya dostupa v oblachnyh strukturah // Odinnadcataya Mezhdunarodnaya Aziatskaya SHkolaseminar Problemy optimizacii slozhnyh sistem Kyrgyzskaya Respublika, IssykKul’skaya oblast’, g. CHolponAta, 27 iyulya — 7 avgusta 2015 g. P. 148–152.
7. Biyashev R. G., Kalimoldaev M. N., Rog O. A. Logicheskij podhod k organizacii mnogokriterial’nogo atributnogo razgranicheniya dostupa // Sovmestnyj vypusk po materialam mezhdunarodnoj nauchnoj konferencii Vychislitel’nye i informacionnye tekhnologii v nauke, tekhnike i obrazovanii (CITech2015) (24–27 sentyabrya 2015 g.) Vychislitel’nye tekhnologii. T. 20, Vestnik KazNU im.Al’Farabi, seriya Matematika, mekhanika i informatika. N 3 (86) CH. 1. P. 275–278.
9. Biyashev R. G., Kalimoldaev M. N., Rog O. A. Predstavlenie ogranichenij modelej atributnogo razgranicheniya dostupa // Izvestiya NAN RK. Seriya fizikomatematicheskaya. 2016. N 1. P. 58–65.
Bibliographic reference: Biyashev R.G., Kalimoldayev M.N., Rog O.A. Typed attributebased access control semantics modeling //journal “Problems of informatics”. 2017, № 1. P. 2537.
Khairetdinov M.S.*, Poller B.V.**, Britvin A.V.**, Sedukhina G. F.*
Novosibirsk State Technical University, 630073, Novosibirsk, Russia
*Institute of Computational Mathematics and Mathematical Geophysics SB RAS, 630090, Novosibirsk, Russia
**Institute of Laser Physics SB RAS, 630090, Novosibirsk, Russia
ACOUSTICOPTICAL SYSTEM OF ULTRALOW FREQUENCIES
UDC 534:621.382
Studying of questions acousticoptical interactions have old history and they are connected, basically, with researches of diffraction of light in ultrasonic and radiofrequency ranges. Questions acousticoptical interactions in area on ultralow practically not investigated. It is in many respects caused by absence of special acoustic radiators in this range of frequencies for carrying out of regular researches. In too time today there are seismic vibrators which are capable along with seismic oscillations in the earth to radiate on ultralow frequencies acoustic oscillations in atmosphere. Centrifugal vibrators CV100 and CV40, in particular, concern them. Acoustic oscillations from them are registered on removals to hundred kilometers and more. It opens possibilities of carrying out of regular experimental researches on acousticoptical interactions in area ultralow frequencies. Authors of the present work offer an original method for study of acousticoptical interactions on ultralow frequencies with powerful seismic vibrators as sources of acoustic oscillations. Along with the generaltheoretical importance of such researches there is a great demand for them in connection with the decision of some the important applied problems, in particular, an acoustic location, the security alarm system, etc.
The description developed by author's acousticoptical systems as a part of vibrator CV40, the optical stand with a laser radiator capacity to 6 Вт in the length of wave 850930 nanometers, frequency of repetition of impulses 1 kHz, the complete set of measuring acoustic stations and a meteorological station is resulted. Results of the executed modeling and natural experiments in which acoustic and optical oscillations were simultaneously registered are presented. Registration of acoustic oscillations is carried out along a line of distribution of a measuring laser beam on base 302 m, removed from a source of acoustic oscillations of vibrator CV40 on 920 m. By means of the ultra resolving spectral analysis processing of the received data is carried out.
The noise immunity of reception of discrete frequency signals from vibrator CV40 in a range of 8.510.5 Hz and also density of noise distribution are estimated. The smaller noise immunity of optical reception in comparison with direct acoustic reception is revealed in 2.5 times. The density of distribution of noise in the optical channel is approximately described by the normal law.
The further development of works is connected with application of a twobeam laser measuring line and use of phase methods of detection for increase of sensitivity of reception of acoustic fluctuations by means of a laser measuring line.
Key words: seismic vibrators, optical stand, natural experiments, results of data processing.
2. Damon R., Maloney W., McMahon D. Interaction of light with ultrasound: the phenomenon and application. Physical Acoustics. Vol. 7, 1970.
3. Alekseev A. S., Glinsky B. M., Kovalevsky V. V., Khairetdinov M. S. et al. Active seismology with powerful vibration sources. Branch Geo of SB RAS Publ. House, Novosibirsk, 2004.
Bibliographic reference: Khairetdinov M.S., Poller B.V., Britvin A.V., Sedukhina G.F. Acousticoptical system of ultralow frequencies //journal “Problems of informatics”. 2017, № 1. P. 3851.
Kholmonov S.M.
Samarkand State University, 140104, Samarkand, Uzbekistan
OPTIMIZATION OF DATA PROCESSING ON THE BASIS OF MODIFIED TRAINING OF NEUTRAL NETWORK AND SEGMENTATION OF CASUAL TIME PROCESSES
UDC 658.512.011
In traditional monitoring systems of industrialtechnological complexes with nonstationary objects of control the quality of casual time series (CTS) identification and data accuracy increase up to the required level is ensured by involving difficult analytical relationships defined in the form of differential, difference, recurrence equations, which are realized by high iterative algorithms. Optimization of identification and data processing is ensured by adjustment parameters of CTS identification model using difficult recurrence relationships when conditions provide availability of extensive prior knowledge.
The perspective approach to elimination of highiterative method of calculation under a priori insufficient of data and uncertainty during developing mechanisms of regulatory and adjustment of nonstationary objects parameters is to develop methods and models that combine opportunity, properties and features of neural networks (NN) with mechanisms for extraction and use of useful knowledge, hidden properties, specific characteristics of CTS. In this case the mathematical apparatus of soft computing successfully completes composition of known methods and algorithms, constructed on the basis of statistical and dynamic approaches, and also expands the capacity of existing toolkits. The important task of NN
using is training of the network, and known algorithms of forward and backpropagation of errors are based on methods of least squares, gradient optimization and their modifications that lead to labor intensive computing of precise values of the optimization functional, especially when it is necessary to solve tasks with high dimensionality.
In this work the mathematical model of data processing optimization is formalized in view of application the mechanism of adjustment, where the task of optimization of data processing is to regulate the value of output result of identification $u(\cdot)$ on the basis of selecting the suboptimal set of NN training parameters.
Concept of probability, evolutionary computation, adaptation of parameters of computational schemes of NN structural components are used for improvement and development methods to optimization data processing and solution of tasks of search during NN training. Solution of designing of adapted modified computational schemes of NN structural components is to form a suboptimal set of parameters such as weights of neurons, synaptic connections coefficients, type of activation function, rational architecture, functional relationships inputs and outputs.
Use of NN radialbasis activation functions gives to network the feature of selftuning and selfadaptation, and allows to carry out dynamic adjustment of parameters of computational schemes of NN structural components. The mechanism for use the methods of parameters adjustment on the basis of selection fractal characteristics, segmentation, filtering and analysis is proposed to empower and improve the adequacy of models to describe objects with nonstationary components of CTS.
The developed program complex includes module of identification and synthesis of dynamic models described nonstationary objects, modules of modified NN training on the basis of radialbasis functions, segmentation, CTS filtering, adjustments of parameters, results of which testing are received at 538 training sets generated from set of CTS measurements.
The study showed that during implementation of developed methods and algorithms for the identification, the number of areas accounted for the change of quasistationary segments increased on 1012%, and the number of accounted classes of nonstationary events increases on 2030% due to determination and formation of suboptimal sets, application of segmentation mechanisms, nonlinear filtering and NN training, and this undoubtedly proves obvious of optimization of data processing based on synthesis of statistical, dynamical and neuro networking models.
Key words: nonstationary object, data processing, identification, optimization, training of neural network, adjustment of parameters, segmentation, nonlinear filtration.
2. Neyrosetevaya model analiza texnollogicheskix vremennix ryadov v ramkax metodologii Data Mining [Neuronetworking model for analysis of technological time series in methodology Data Mining] / A. M. Vulfin, A. I. Frid // Informasionnoupravlyayushiye sistemi. 2011. N 5. P. 31–38.
3. Scott, A. C. Neuroscience: a mathematical primer / Alan S. Scott. — New York: Springer–Verlag, 2002.
4. Jumanov I. I., Kholmonov S. M. Optimizasiya obucheniya neyrosetevoy sistemi obrabotki dannix na osnove statisticheskix svoystv informasii [Optimization of training of neuronetworking system of data processing on the basis of information statistical properties] // Problemi informatiki I energetiki. 2011. N 3. P. 50–56.
5. Sezer, E. Employment and comparison of different artificial neural networks for epilepsy diagnosis from EEG signals / E. Sezer, H. Isik, E. Saracoglu // J. Med. Syst. 2012. N 36. Ð. 347.
6. Chipiga A. F., Voronkin R. A. Obucheniye iskusstvennix neyronnix setey putyom sovmestnogo ispolzovaniya metodov lokalnoy optimizasii I geneeticheskix algoritmov [Training of artificial neural networks by sharing methods of local optimization and genetic algorithms] // Izvestiya TRTU. T. 33. N 4. P. 172–174.
7. Lawhern, V. Detection and classification of subjectgenerated artifacts in EEG signals using autoregressive models / V. Lawhern, W. D. Hairston, K. McDowell, M.Westerfield, K. Robbins // Journal of Neuroscience Methods. 2012. N 208. Ð. 181.
8. Jumanov I. I., Abdullayev A. N. Kontrol tochnosti peredachi informasii v sistemax avtomatizasii izmereniya I obrabotki dannix nestasionarnoy prirodi [Control of information accuracy during transfer in systems of automation measurement and processing of data with nonstationary nature] // WCIS — 2006, TGTU, Tashkent. P. 15–21.
9. Djumanov O. I. Adaptive designing for neuronetworking system of processing the data with nonstationary nature // Peerreviewed & Open access journal ATI — Applied Technologies & Innovations, Issue 1, April 2011. Prague, 2011. V. 4. P. 48–57.
10. Ningsheng Gong, Wei Shao, Hongwei Xu. The Conjugate Gradient Method with Neural Network. Control, 2010, IEEE, Nanjing, Jiangsu, China.
11. Yarushkina N. G. Osnovi nechetkix I gibridnix system [Bases of fuzzy and hybrid systems]: Uchebnoye posobiye. M.: Finansi I statistika. 2004. P. 70–120.
12. Khlmonov S. M. Optimizasiya obrabotki dannix nz osnove metodov otbora informativnix priznakov dlya prognozirovaniya nestacionarnix obyektov [Optimization of data processing on the basis of methods to selecting informative attributes for forecasting nonstationary objects] // X naucnoprakticheskaya konferensiya Sovremenniye instrumentalniye sistemi, informasionniye texnologii I innovasii, 19–23 marta 2013 g. Moskva, 2013. P. 264–269.
13. Primeneniye metodov segmentasii k obrabotke geofizicheskix dannix [Application of methods of segmentation to processing the geophysical data] / V. V. Geppner, A. B. Tristanov, P. P. Firstov // Kompleksniye seysmologicheskiye I geofizicheskiye issledovaniya Kamchatki, KSiGIK — 2006. Materiali Vserossiyskoy nauchnotehnicheskoy konferensii, 2006. P. 183–187.
14. Sinha, R. K. Artificial neural networks and wavelet based automated detection of sleep spindles, REM sleep and wake states / R. K. Sinha // J. Med. Syst. 2008. N 32. P. 291.
Bibliographic reference: Kholmonov S.M. Optimization of data processing on the basis of modified training of neutral network and segmentation of casual time processes //journal “Problems of informatics”. 2017, № 1. P. 5261.
Moiseenko V.V., Rodionov A.S.
Institute of Computational Mathematics and Mathematical Geophysics of Siberian Branch of Russian Academy of Sciences, 630090, Novosibirsk, Russia
AGE DEPENDENCY OF COAUTHORSHIP IN SCIENTIFIC RESEARCHES IN ACADEMIA
UDC 510519.24
The paper is devoted to examination of coauthorship of researchers from different age groups in papers prepared in an institute of natural sciences. Research bases on example of real statistics of a special institute using databases that include personnel and productive indices. We show that number and structure of coauthors depends on a scientist's age. The dynamics of number of coauthors per paper during 20012015years is shown also.
We consider the following indices:
1) part of researchers who have publications made in coauthorship;
2) average number of coauthors per researcher in each age group;
3) average number of coauthors per paper for a researcher in each age group;
4) average number of coauthors per one researcher in given age group in the case of coauthorship with researchers from other age groups;
5) average number of coauthors per paper for a researcher in given age group in the case of coauthorship with researchers from other age groups.
The following main conclusions are done based on our data:
1) Most active in coauthorship are researcher in the 1\textsuperscript{st} age group (up to 35 years old). This may be explained by the fact that young researchers usually help their supervisors and so good part of ideas and problem statements in the papers belongs to coauthors from older groups. Next peak of coauthorship corresponds just to oldest (5th) group. Researcher in this group rarely are active in personal researches, they mostly lead researches performed by doctoral students and young scientists.
2) We may state that average number of coauthors in papers devoted to applied mathematics is not as large as in social or natural (chemistry, biology, medicine, etc.) sciences, but it is larger when a paper needs numerical experiments, that is programming and executing proposed algorithms, which means including programmers into a team.
3) There is clear trend to increasing number of papers and coauthors in last years. This may be explained by the following consideration: number of publications became main index of a researcher's quality. In this case it is quite natural that instead of publishing one paper per year without coauthorship or with 12 coauthors, researchers agree that it is easier and better for annual reporting publish 35 papers per year with 24 or even more coauthors.
Key words: scientific process, subsystem, age group, coauthorship, average number of coauthors, publishing scientific paper.
2. Karpachev G. I., Moiseenko V. V. Nekotorye voprosy retrospektivnogo analiza struktury i deyatel’nosti nauchnyh kadrov akademicheskogo instituta // Problemy informatiki. 2009. N 1. S. 75–78.
3. Klejn L. Drama soavtorstva // Gazeta Troickij variant. 2015. N 1(82). S.11.
4. Moiseenko V. V. Imitacionnoe modelirovanie nauchnogo processa Akademicheskogo instituta // Problemy informatiki. 2014. N 4. S. 80–87.
5. Sergeev N. M. EHtika soavtorstva i ehtika citirovaniya // Sajt VIVOS VOCO [ehlektronnyj resurs]. Data obrashcheniya 21.05.2016. URL: http://vivovoco.astronet.ru/VV/PAPERS/ECCE/ETHICS/SERG.HTM.
6. Plotnikov V. G. O soavtorstve v nauke // Put’ nauki. Mezhdunarodnyj nauchnyj zhurnal. 2015. N 10 (20). S. 149–151.
7. Titov I. I., Blinov A. A. Issledovanie struktury i ehvolyucii setej nauchnogo soavtorstva na osnove analiza novosibirskih publikacij v oblasti biologii i mediciny // Vavilovskij zhurnal genetiki i selekcii. 2014. T. 18. N 4/2. S. 939–944.
8. Hamermesh D. S. Age, cohort and coauthorship // NBER Working Papers 20938, National Bureau of Economic Research, Inc, 2015.
Bibliographic reference: Moiseenko V.V., Rodionov A.S. Age dependency of coauthorship in scientific researches in academia //journal “Problems of informatics”. 2017, № 1. P. 6273.
Malyshkin* V.E., Perepelkin* V.A., Schukin** G.A.
Institute of Computational Mathematics and Mathematical Geophysics SB RAS, 630090, Novosibirsk, Russia
*Novosibirsk State University, 630090, Novosibirsk, Russia
**Novosibirsk State Technical University, 630073, Novosibirsk, Russia
DISTRIBUTED ALGORITHM FOR DATA ALLOCATION IN LUNA FRAGMENTED PROGRAMMING SYSTEM
UDC 004.021
The paper presents a scalable distributed algorithm for static and dynamic data allocation in LuNA fragmented programming system. The proposed algorithm takes into account data structure of the application numerical model executed, enables static and dynamic load balancing and can be used with various network topologies. Implementation of numerical models on multicomputers with large number of computing nodes is a challenging problem in the domain of highperformance parallel computing. Effective resources allocation and balancing strategies are necessary to achieve good efficiency and scalability of parallel programs. LuNA is a fragmented programming system which is being developed in Supercomputer Software Department of Institute of Computational Mathematics and Mathematical Geophysics SB RAS. LuNA’s main objective is automation of construction of parallel programs, which implement largescale numerical models for largescale multicomputers.
In LuNA system an application algorithm is represented as two sets: a set of immutable data pieces (data fragments, DF) and a set of sideeffect free computational processes (computational fragments, CF). All these fragments can be distributed over computational nodes of a multicomputer. Each DF is produced by one CF and can be used as an input by other CFs; each CF is executed only once and requires values of all its input DFs to be gathered on a single node for its execution. Execution of a program is managed by LuNA's runtime system. The runtime system performs distribution and migration of DFs and CFs over nodes of a multicomputer and deliverance of input DFs to their corresponding CFs to provide execution of all CFs in the program.
There are several steps in processing of DFs and CFs. Each DF must initially be assigned to a node for its storage and be allocated on this node. Each CF must be assigned a node for its execution and also be allocated on the node. Efficiency of LuNA program execution (in terms of running time, memory consumption, communications amount, etc.) heavily depends on CFs and DFs distribution. To achieve suitable efficiency, during execution of a program
fragments may be transferred between nodes in order to equalize workload. Such dynamic load balancing is also performed by the runtime system. The algorithm described in the paper is used as for initial data allocation, as well as for
dynamic load balancing. Although its description is given in respect to data fragments only, in the same way it can also be used (and is used actually) for managing allocation of computational fragments.
First, a mapping of data fragments to a range of integer indices must be constructed. The mapping is an input for the algorithm described. The mapping is expected to map dependent neighboring data fragments on the same or close indices. Construction of such a mapping is a complex task, which is out of the scope of the paper. Construction of a suitable mapping, a knowledge of algorithm structure is required, human help is expected. To create a mapping of a multidimensional data structures to the onedimensional index range spacefilling curves (such as Hilbert spacefilling curve) may be used to preserve data neighborhood. The mapping is a constant function during execution of a program, thus it can
be computed on any node without communications.
Second, the given indices range is split into segments with number of segments equal to the number of computational nodes. Each node is given its own segment, neighboring nodes in a line topology receiving neighboring segments. Each data fragment will be allocated on the node containing index of its mapping. For that, assuming the data fragment was produced on some node, first its mapping is computed and then, if necessary, the fragment is
transferred across neighboring nodes until its final node is found. Because indices are ordered and for each node its neighbors are known, direction of fragments' movement in the line of nodes can always be determined at any node. In such way data allocation requires only local communications between neighboring nodes.
For the sake of dynamic load balancing the mapping of segments of indices to nodes can be changed. First, for each index in a segment its workload is evaluated. Formulas for workload evaluation can be different, for example value of workload can be proportional to a volume of memory occupied by data on the node. Node's workload is computed as sum of its indices workloads. Second, workloads of neighboring nodes are compared periodically. If the workload difference exceeds a threshold, load rebalancing will be performed. For that neighboring nodes shift common border of their segments (from overloaded node to underloaded one), causing DFs mapped to migrate thus equalizing workload. Because order of indices is preserved during border shift, data locality is preserved too. Load balancing algorithm is distributed (executes on each node) and uses only local communications.
Usage of only local communications for data allocation and load balancing makes proposed algorithm scalable to a large number of computational nodes, because no global synchronization (that can impede scalability) is used.
Testing of the algorithm has shown its potential for scalability; performance of fragmented program for problem used was just several times worse than of usual MPI implementation for the same problem, given that MPI implementation with dynamic load balancing takes more time and effort to program instead of LuNA implementation, where data allocation and load balancing are implemented automatically.
Key words: scalable distributed system algorithm, dynamic data allocation, distributed algorithms with local interactions, fragmented programming technology, fragmented programming system LuNA.
2. Malyshkin, V. E., Perepelkin, V. A. Optimization Methods of Parallel Execution of Numerical Programs in the LuNA Fragmented Programming System // J. Supercomputing. 2012. V. 61, N 1, P. 235–248.
3. Malyshkin, V. E., Perepelkin V. A. The PIC Implementation in LuNA System of Fragmented Programming // J. Supercomputing. 2014. V. 69. N 1. P. 89–97.
4. Kraeva, M. A., Malyshkin, V. E. Assembly Technology for Parallel Realization of Numerical Models on MIMDMulticomputers // J. Future Generation Computer Systems. 2001. V. 17. N 6. P. 755–765.
5. Kireev, S. E., Malyshkin V. E. Fragmentation of Numerical Algorithms for Parallel Subroutines Library // J. Supercomputing. 2011. V. 57. N 2. P. 161–171.
6. Kraeva, M. A., Malyshkin, V. E. Algoritmy dinamicheskoi balansirovki zagruzki pri realizaczii metoda chasticz v yacheikakh na MIMDmultikomputerakh. J. Programmirovanie. 1999. N 1. P. 47–53.
7. Hu, Y. F., Blake, R. J. An Improved Diffusion Algorithm for Dynamic Load Balancing. J. Parallel Computing. 1999. V. 25. N 4. P. 417–444.
8. Corradi, A., Leonardi, L., Zambonelli F. Performance Comparison of Load Balancing Policies Based on a Diffusion Scheme / EuroPar’97 Parallel Processing. 1997. P. 882–886. Springer, Heidelberg.
9. Anderson, J. M., Lam, M. S. Global Optimizations for Parallelism and Locality on Scalable Parallel Machines / ACMSIGPLAN PLDI’93. 1993. P. 112–125. ACM New York, USA.
10. Li, J., Chen, M. The Data Alignment Phase in Compiling Programs for DistributedMemory Machines // J. Parallel and Distributed Computing. 1991. V. 13. N 2. P. 213–221.
11. Lee P. Efficient Algorithms for Data Distribution on Distributed Memory Parallel Computers // J. IEEE Transactions on Parallel and Distributed Systems. 1997. V. 8. N 8. P. 825–839.
12. YuKwong Kwok, Ahmad, I. Design and Evaluation of Data Allocation Algorithms for Distributed Multimedia Database Systems // IEEE Journal on Selected Areas in Communications. 1997. V. 14. N 7. P. 1332–1348.
13. Iacob, N. M. Fragmentation and Data Allocation in the Distributed Environments // Annals of the University of Craiova — Mathematics and Computer Science Series. 2011. V. 38. N 3. P. 76–83.
14. Jagannatha, S., Geetha, D. E., Suresh Kumar, T. V., Rajani Kanth, K. Load Balancing in Distributed Database System using Resource Allocation Approach // J. Advanced Research in Computer and Communication Engineering. 2013. V. 2. N 7. P. 2529–2535.
15. Honicky, R. J., Miller E. L. Replication Under Scalable Hashing: A Family of Algorithms for Scalable Decentralized Data Distribution / 18th International Parallel and Distributed Processing Symposium, 2004.
16. Alicherry, M., Lakshman, T. V. Network Aware Resource Allocation in Distributed Clouds / INFOCOM 2012. P. 963–971.
17. AuYoung, A., Chun, B. N., Snoeren, A. C., Vahdat, A. Resource Allocation in Federated Distributed Computing Infrastructures // First Workshop on Operating System and Architectural Support for the Ondemand IT InfraStructure, 2004.
18. Raman, R., Livny M., Solomon, M. Matchmaking: Distributed Resource Management for High Throughput Computing // J. Cluster Computing. 1999. V. 2. N 1. P. 129–138.
19. Reddy, C., Bondfhugula, U. Effective Automatic Computation Placement and Data Allocation for Parallelization of Regular Programs / 28th ACM International Conference on Supercomputing. 2014. P. 13–22. ACM New York, USA.
20. Baden, S. B., Shalit, D. Performance Tradeoffs in Multitier Formulation of a Finite Difference Method // ICCS 2001, LNCS. V. 2073. P. 785–794. Springer, Heidelberg.
21. Kenichiro Ishikawa. ASURA: Scalable and Uniform Data Distribution Algorithm for Storage Clusters. Computing Research Repository, abs/1309.7720, 2013.
22. Chawla, A., Reed B., Juhnke, K., Syed, G. Semantics of Caching with SPOCA: A Stateless, Proportional, OptimallyConsistent Addressing Algorithm // USENIX Annual Technical Conference 2011. P. 33–33. USENIX Association.
23. Lowder, J. K., King, P. J. H. Using SpaceFilling Curves for Multidimensional Indexing // Advances in Databases. LNCS. V. 1832. P. 20–35. Springer, Heidelberg, 2000.
24. Moon, B., Jagadish, H. V., Faloutsos, C., Saltz, J. H. Analysis of the Clustering Properties of the Hilbert SpaceFilling Curve // J. IEEE Transactions on Knowledge and Data Engineering. 2000. V. 13. N 1. P. 124–141.
Bibliographic reference: Malyshkin V.E., Perepelkin V.A., Schukin G.A. Distributed algorithm for data allocation in luna fragmented programming system //journal “Problems of informatics”. 2017, № 1. P. 7488.