Институт проблем информатики Российской Академии наук
Институт проблем информатики Российской Академии наук
Российская Академия наук

Институт проблем информатики Российской Академии наук




«INFORMATICS AND ITS APPLICATIONS»
Scientific journal
Volume 7, Issue 2, 2013

Content | Bibliography | About  Authors

Abstract and Keywords.

PARAMETRICAL STATISTICAL AND ANALYTICAL MODELING OF DISTRIBUTIONS IN NONLINEAR STOCHASTIC SYSTEMS ON MANIFOLDS .

  • I.N. Sinitsyn   IPI RAN, sinitsin@dol.ru

Abstract: Discrete parametrical statistical and analytical modeling methods in nonlinear Ito stochastic systems on manifolds with Wiener and Poisson noises have been developed. For one- and multidimensional densities parametrization, the coefficients in orthogonal expansions of different orders are taken. Special attention is paid to nonlinear correlational theory of statistical and analytical modeling.

Keywords:  analytical modeling; method of normal approximation; nonlinear correlational theory; nonlinear Ito stochastic system on manifold; orthogonal expansions method; parametrization of one- and multidimensional densities; statistical linearization method; statistical modeling

EVALUATION METHODS FOR EFFICIENCY AND DIRECTIVE TERMS OF PERFORMANCE OF RESOURCE-INTENSIVE COMPUTING TASKS.

  • I. K. Kupalov-Yaropolk   Lebedev Institute of PrecisionMechanics and Computer Engineering, Russian Academy of Sciences, kupyar@rambler.ru
  • Yu. E. Malashenko  Dorodnicyn Computing Center, Russian Academy of Sciences, malash09@ccas.ru
  • I. A. Nazarova   Dorodnicyn Computing Center, Russian Academy of Sciences, irina-nazar@yandex.ru
  • A. F. Ronzhin   Dorodnicyn Computing Center, Russian Academy of Sciences, raf-zao-zt@yandex.ru

Abstract: The problem of effective use of the heterogeneous computing system is considered at parallel processing of diverse tasks. In a case of date violation of works completion, the processor spent time belongs to production losses. Planning and optimization controls are exercised on the basis of the guaranteed estimates constructed for the worst case.

Keywords:  parallel computing; multiprocessor systems; optimization; the principle of guaranteed result

ON ESTIMATION OF THE EFFECTIVE BANDWIDTHS IN A SYSTEM WITH REGENERATIVE INPUT.

  • A. V. Borodina   Institute of Applied Mathematical Research, Karelian Research Center, Russian Academy of Sciences, borodina@krc.karelia.ru
  • E. V. Morozov   Institute of Applied Mathematical Research, Karelian Research Center, Russian Academy of Sciences, emorozov@karelia.ru

Abstract: The effective bandwidths (EB) of a communication system are considered. The EB guarantees that the stationary workload overflow/loss probability to exceed a threshold is limited by a (small) quantity. It is shown how to calculate EB in a fluid queue fed by an input with the independent increments. Then, a fluid system with regenerative input is considered. Using heuristic arguments, an approximation of the limiting logarithmic exponential moment generating function of the input was deduced. Numerical simulations show a satisfactory accuracy of the approximation applied to a few systems with regenerative input.

Keywords:  effective bandwidths; fluid queue; workload process; overflow/loss probability; effective bandwidths approximation; regenerative estimation

STATIONARY WAITING TIME DISTRIBUTION IN QUEUEING SYSTEM WITH NEGATIVE CUSTOMERS AND BUNKER FOR OUSTED CUSTOMERS UNDER FIRST–FIFO–FIFO SERVICE DISCIPLINE.

  • R. V. Razumchik  IPI RAN, rrazumchik@ieee.org

Abstract: Consideration is given to the single server queueing system with Poisson input flows of ordinary and negative customers. An arriving ordinary customer occupies one place in an infinite buffer. Negative customer upon arrival pushes out one ordinary customer from the queue in the buffer to another queue (bunker) and leaves the system. Customers frombunker are served with relative priority. Service times of customers from buffer and bunker are both exponentially distributed but with different rates. It is assumed that negative customer always pushes out the first customer in the queue in the buffer and after service completion the first customer in the queue in the buffer enters server or, if buffer is empty, the first customer in the queue in the bunker. Stationary waiting time distribution of an arriving ordinary customer in terms of Laplace–Stieltjes transform is obtained.

Keywords:  queueing system; negative customers; waiting time

CENTRAL LIMIT THEOREM FOR GENERALIZED CROSS-VALIDATION FUNCTION IN WAVELET THRESHOLDING METHOD.

  • O. V. Shestakov   Department of Mathematical Statistics, Faculty of Computational Mathematics and Cybernetics, M. V. Lomonosov Moscow State University; IPI RAN, oshestakov@cs.msu.su

Abstract: In this paper the asymptotic properties of generalized cross-validation function in wavelet thresholding method are analyzed. Generalized cross-validation function is minimized to choose the threshold. Asymptotic normality of generalized cross-validation function is proved.

Keywords:  thresholding; generalized cross-validation; adaptive threshold; unbiased risk estimate; asymptotic normality

STATISTICAL TESTING FOR THE NONEXECUTABILITY OF FRAGMENTS OF THE CODE OF A LINEAR PROGRAM.

  • V. Yu. Korolev   Faculty of Computational Mathematics and Cybernetics, M.V. Lomonosov Moscow State University; IPI RAN, victoryukorolev@yandex.ru
  • R. L. Smelyansky   Faculty of Computational Mathematics and Cybernetics, M.V. Lomonosov Moscow State University, smel@sc.msu.ru
  • T. R. Smelyansky   Faculty of Computational Mathematics and Cybernetics, M.V. Lomonosov Moscow State University, smelyanskiy.t@bk.ru
  • A. V. Shalimov   Faculty of Computational Mathematics and Cybernetics, M.V. Lomonosov Moscow State University, ashalimov@lvk.cs.msu.su

Abstract: A problem of statistical testing for the nonexecutiveness of fragments of the code of a linear program is considered. Methods based on the minimization of prior error probabilities are considered as well as those based on the minimization of posterior error probabilities.

Keywords: testing statistical hypotheses; geometric distribution; probability of the error of the first kind; probability of the error of the second kind; Neyman–Pearson lemma; posterior error probability

BAYESIAN RECURRENT MODEL OF RELIABILITY GROWTH:
UNIFORM DISTRIBUTION OF PARAMETERS.

  • A. A. Kudriavtsev   Faculty of Computational Mathematics and Cybernetics, M. V. LomonosovMoscow State University, nubigena@hotmail.com
  • I. A. Sokolov   IPI RAN, isokolov@ipiran.ru
  • S. Ya. Shorgin   IPI RAN, sshorgin@ipiran.ru

Abstract: The paper is devoted to justification of expediency of Bayesian approach at the solution of the tasks connected with determination of reliability of complex modifiable systems. As illustration, average value of reliability of system is presented in a case where characteristics of “defectiveness” and “efficiency” of the means correcting imperfections of system are uniformly distributed.

Keywords:  modifiable information systems; reliability theory; Bayesian approach

THE ERROR-IN-VARIABLES MODEL IDENTIFICATION ON THE BASIS OF DEMING’S APPROACH.

  • V. S. Timofeev   Novosibirsk State Technical University, netsc@rambler.ru
  • V. Yu. Schekoldin   Novosibirsk State Technical University, raix@ngs.ru
  • A. Yu. Timofeeva  Novosibirsk State Technical University, supernasty@mail.ru

Abstract: Some approaches to regression models constructing with stochasticity for both endogenous and exogenous variables are considered. The original geometric interpretation of particular cases of Deming regression for parameter estimation’s functional is suggested. The proposition of mutual arrangement for straight, inverse, diagonal, and orthogonal regressions is proved. The bias and standard deviation for regression parameter estimators in dependence of weight’s coefficients ratio in Deming model has been obtained.

Keywords:  least square estimation; Deming regression; geometric interpretation; dispersion ellipse

ASYMPTOTIC NORMALITY OF THE ESTIMATION OF THE MULTIVARIATE LOGISTIC REGRESSION.

  • A. Yu. Khaplanov   M.V. Lomonosov Moscow State University, Khaplanova@gmail.com

Abstract: Estimation of characteristics of the multivariate logistic regression with a diverging number of covariates has been made. The convergence rate for estimate of characteristics of the multivariate logistic regression coefficients has been obtained. Asymptotic normality of its rejection has been proved.

Keywords:  logistic regression; convergence rate; asymptotic normality; high-dimensional covariates

ASYMPTOTIC EXPANSIONS FOR THE DISTRIBUTION FUNCTIONS OF STATISTICS CONSTRUCTED FROM SAMPLES WITH RANDOM SIZES.

  • V. E. Bening   Department o fMathematical Statistics, Faculty of Computational Mathematics and Cybernetics, M. V. Lomonosov Moscow State University; IPI RAN, bening@yandex.ru
  • N. K. Galieva   Kazakhstan Branch of the M. V. Lomonosov Moscow State University, nurgul u@mail.ru
  • V. Yu. Korolev   Faculty of Computational Mathematics and Cybernetics, M. V. LomonosovMoscow State University; IPI RAN, victoryukorolev@yandex.ru

Abstract: A general transfer teorem is proved making it possible to construct asymptotic expansions for the distribution function of a statistic constructed from the sample with a random size from the asymptotic expansion for the distribution function of the random sample size and the asymptotic expansion for the distribution function of the same statistic constructed from the samples with a nonrandom size.

Keywords:  sample with a random size; asymptotic expansion; transfer theorem; mixture of probability distributions; Laplace distribution; Student distribution

ON CONVERGENCE OF RANDOM WALKS GENERATED BY COMPOUND COX PROCESSES TO LEVY PROCESSES.

  • V. Yu. Korolev   Faculty of Computational Mathematics and Cybernetics, M.V. Lomonosov Moscow State University; IPI RAN, victoryukorolev@yandex.ru
  • L. M. Zaks  Department of Modeling and Mathematical Statistics, Alpha-Bank, lily.zaks@gmail.com
  • A. I. Zeifman   Vologda State Pedagogical University; IPI RAN, a_zeifman@mail.ru

Abstract: A functional limit theoremis proved establishing weak convergence of random walks generated by compound doubly stochastic Poisson processes to Levy processes in the Skorokhod space. As corollaries, theorems on convergence of random walks with jumps having finite variances to Levy processes with mixed normal distributions, in particular, to stable Levy processes have been proved.

Keywords:  stable distribution; Levy process; stable Levy process; compound doubly stochastic Poisson process (compound Cox process); Skorokhod space; transfer theorem

STATISTICAL MECHANISMS OF THE SUBJECT DOMAINS ASSOCIATIVE PORTRAITS FORMATION ON THE BASIS OF BIG NATURAL LANGUAGE TEXTS FOR THE SYSTEMS OF KNOWLEDGE EXTRACTION.

  • M. M. Charnine   IPI RAN, 1@keywen.com
  • N. V. Somin   IPI RAN, somin@post.ru
  • I. P. Kuznetsov   IPI RAN, igor-kuz@mtu-net.ru
  • Yu. I. Morozova  IPI RAN, judez@yandex.ru
  • I. V. Galina   IPI RAN, irn gl@mail.ru
  • E. B. Kozerenko   IPI RAN, kozerenko@mail.ru

Abstract: Associative relations between terms, concepts and other elements of natural language play an important role in decision of a wide variety of application tasks including intelligent texts processing, knowledge extraction, and management comprizing the formation of knowledge bases and semantic information retrieval. The paper presents the methods of automatic establishment of the associative relations between terms and concepts in the texts from Internet and creation of subject domains associative portraits designed for the tasks of intelligent systems development. An associative portrait of a subject domain (APSD) is a dictionary of the meaningful support terms and word combinations interconnected by associative relations. It is essential that the APSD are constructed automatically on the basis of statistical analysis of big volumes of texts. The theoretical impact of the proposed method consists in the use of statistics, corpus linguistics, and distributional semantics for processing big volumes of natural language texts which are dynamically updated and enriched in the Internet for constructing the model of a subject domain in the form of APSD.

Keywords:  automatic processing of text corpora; statistical methods; intelligent Internet technologies; lexical semantic analysis; knowledge extraction from texts; semantic retrieval; semantic vectors; semantic context space

INFORMATION TECHNOLOGIES FOR CREATING THE DATABASE OF EQUIVALENT VERBAL FORMS IN THE RUSSIAN-FRENCH MULTIVARIANT PARALLEL CORPUS.

  • S. Loiseau  Universite Paris 13, Sorbonne Paris Cite, Laboratoire LDI (Lexiques, dictionnaires, informatique), CNRS, UMR 7187, sylvain.loiseau@univ-paris13.fr
  • D. V. Sitchinava  Institute of the Russian Language of the Russian Academy of Sciences, mitrius@gmail.com
  • A. A. Zalizniak  Institute of Linguistics of the Russian Academy of Sciences; Institute of Informatics Problems of the Russian Academy of Sciences, anna.zalizniak@gmail.com
  • I. M. Zatsman  Institute of Informatics Problems of the Russian Academy of Sciences, iz_ipi@a170.ipi.ac.ru

Abstract: The Russian-French parallel corpus as a part of the Russian National Corpus is being transformed into a multivariant corpus with several translations corresponding to each original texts. Concurrently, a database of functionally equivalent lexicogrammatical verbal forms is being created using the multivariant corpus. The main purpose of database creation is to calculate the statistical estimates of the equivalences between Russian and French verbal forms. The paper discusses an information technology for creating the Russian-French multivariant parallel corpus and the database simultaneously.

Keywords:  parallel multivariant corpora; Russian National Corpus; information technologies; XML marking up Russian-French parallel texts; lexicogrammatical form; functional equivalence; statistical estimates of equivalences