Advanced search in
Research products
arrow_drop_down
Searching FieldsTerms
Any field
arrow_drop_down
includes
arrow_drop_down
Include:
26 Research products, page 1 of 3

  • Publications
  • Research software
  • 2012-2021
  • Open Access
  • Russian
  • Informatika
  • Mémoires en Sciences de l'Information et de la Communication
  • Hal-Diderot
  • Hyper Article en Ligne

10
arrow_drop_down
Relevance
arrow_drop_down
  • Open Access Russian
    Authors: 
    D. I. Kachkou;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The article is an essay on the development of technologies for natural language processing, which formed the basis of BERT (Bidirectional Encoder Representations from Transformers), a language model from Google, showing high results on the whole class of problems associated with the understanding of natural language. Two key ideas implemented in BERT are knowledge transfer and attention mechanism. The model is designed to solve two problems on a large unlabeled data set and can reuse the identified language patterns for effective learning for a specific text processing problem. Architecture Transformer is based on the attention mechanism, i.e. it involves evaluation of relationships between input data tokens. In addition, the article notes strengths and weaknesses of BERT and the directions for further model improvement.

  • Open Access Russian
    Authors: 
    Rykhard Bohush; I. Yu. Zakharava; Sergey Ablameyko;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    In the paper the algorithm for object detection in high resolution images is proposed. The approach uses multiscale image representation followed by block processing with the overlapping value. For each block the object detection with convolutional neural network was performed. Number of pyramid layers is limited by the Convolutional Neural Network layer size and input image resolution. Overlapping blocks splitting to improve the classification and detection accuracy is performed on each layer of pyramid except the highest one. Detected areas are merged into one if they have high overlapping value and the same class. Experimental results for the algorithm are presented in the paper.

  • Open Access Russian
    Authors: 
    Yu. N. Sotskov; Natalja G. Egorova; Natalja M. Matsveichuk;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    Time-management refers to planning the available time in correspondence with personal goals and preferences of the employee. The criterion for efficient time-management lies in achieving definite goals in minimum possible time. This paper presents a short review of the literature on time-management. It is shown how scheduling theory may be used for optimal planning in time-management. The problem of minimizing the total (average) weighted completion time of the planned jobs by an employee is considered provided that only lower and upper bounds of the possible processing time of each job are known before scheduling. Software for constructing a permutation of the given jobs with the largest relative semi-perimeter of the optimality parallelepiped has been developed. Computational experiment on the computer showed the effectiveness of the developed algorithms for constructing schedules for time-management.

  • Open Access Russian
    Authors: 
    M. L. Radziukevich; V. F. Golikov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The main options for the formation of a shared secret using synchronized artificial neural networks and possible patterns of behavior of a cryptanalyst are considered. To solve the problem of increasing the confidentiality of the generated shared secret, if it is used as a cryptographic key, it is proposed to use the mixing a certain number of results of individual synchronizations (convolution). As a mixing function, we consider the convolution of the vectors of network weights by bitwise addition modulo 2 of all the results of individual synchronizations. It is shown that the probability of success of a cryptanalyst is reduced exponentially with an increase of the number of terms in the convolution and can be chosen arbitrarily small. Moreover, the distribution law of the generated key after convolution is close to uniform and the uniformity increases with the number of terms in the convolution.

  • Open Access Russian
    Authors: 
    A. Y. Kharin;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    In the problems of data flows analysis, the problems of statistical decision making on parameters of observed data flows are important. For their solution it is proposed to use sequential statistical decision rules. The rules are constructed for three models of observation flows: sequence of independent homogeneous observations; sequence of observations forming a time series with a trend; sequence of dependent observations forming a homogeneous Markov chain. For each case the situation is considered, where the model describes the observed stochastic data with a distortion. "Outliers" ("contamination") are used as the admissible distortions that adequately describe the majority of situations appear in practice. For such situations the families of sequential decision rules are proposed, and robust decision rules are constructed that allow to reduce influence of distortion to the efficiency characteristics. The results of computer experiments are given to illustrate the constructed decision rules.

  • Open Access Russian
    Authors: 
    N. N. Guschinsky; O. Battaia; A. Dolgui;
    Publisher: HAL CCSD
    Country: France

    A problem of design of a flow line composed of stationary machines for sequential batch machining is considered. A MIP model for synthesis of manufacturing process structure is proposed. The model takes into account main technological and structural constraints with the criterion of the minimal line equipment cost. Results of solving a real industrial example are also presented.

  • Open Access Russian
    Authors: 
    Y. A. Drybin; S. V. Sadau; V. S. Sadau;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    It is shown that the choice of the time sampling parameter of the digital model of a continuous dynamic system with chaotic modes based on its dynamics makes it possible to control the characteristics of the output sequence, including avoiding short cycles and periodic behavior modes. On the example of the Lorentz system, the analysis of the law of motion of a chaotic system, linearized in the vicinity of points of stable and unstable equilibrium, is carried out, on the basis of which the parameters of the mathematical model of the generator of pseudo-random numbers are selected. The output sequence of numbers generated in proposed way is subjected to statistical and correlation analysis. Based on the results of the tests carried out, we can say that the obtained pseudo-random sequences based on continuous chaotic systems have statistically random properties and can be used in steganographic and cryptographic systems.

  • Open Access Russian
    Authors: 
    Yu. V. Pottosin;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The orthogonal disjunctive normal forms (DNFs) of Boolean functions have wide applications in the logical design of discrete devices. The problem of DNF orthogonalization is to get for a given function such a DNF that any two its terms would be orthogonal, i. e. the conjunction of them would be equal identically to zero. An approach to solve the problem using the means of graph theory is suggested. The approach is proposed by representation of the function as perfect DNF. Obtaining all the intervals of the Boolean space where the given function has value 1 is supposed, and the intersection graph of those intervals is considered. Two methods to obtain a minimum orthogonal DNF are considered. One of them reduces the problem toward finding out the smallest dominating set in the graph by covering its vertices with their closed neighborhoods, the other - to obtain the maximum independent set by lexicographic enumeration. It is shown how the suggested approach can be extended on incompletely specified Boolean functions.

  • Open Access Russian
    Authors: 
    R. S. Vashkevich; E. S. Azarov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The paper investigates the problem of voice activity detection from a noisy sound signal. An extremely compact convolutional neural network is proposed. The model has only 385 trainable parameters. Proposed model doesn’t require a lot of computational resources that allows to use it as part of the “internet of things” concept for compact low power devices. At the same time the model provides state of the art results in voice activity detection in terms of detection accuracy. The properties of the model are achieved by using a special convolutional layer that considers the harmonic structure of vocal speech. This layer also eliminates redundancy of the model because it has invariance to changes of fundamental frequency. The model performance is evaluated in various noise conditions with different signal-to-noise ratios. The results show that the proposed model provides higher accuracy compared to voice activity detection model from the WebRTC framework by Google.

  • Open Access Russian
    Authors: 
    Vladimir Sapozhnikov; Vl.V. Sapozhnikov; D. V. Efanov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    Simpler than known structure of the system with error correction in calculations is proposed based on duplication and triplication of blocks with majority principle of choosing the values of signals. It is advisable to use the new fault-tolerant structure for automation devices with combinational logic. In fault-tolerant structure synthesis, the parity method is used to establish the fact of a fault in the main logic unit and the logical complement method is used determine incorrectly calculated output functions and to generate signals for their correction. The method also allows to adjust the values of incorrectly calculated functions. Structural diagram and description of error correction system are given. The synthesis algorithm of control equipment is described with minimization of the technical implementation complexity. The experiment results with control combinational circuits are given, confirming the high efficiency of proposed system structure with error correction.

Advanced search in
Research products
arrow_drop_down
Searching FieldsTerms
Any field
arrow_drop_down
includes
arrow_drop_down
Include:
26 Research products, page 1 of 3
  • Open Access Russian
    Authors: 
    D. I. Kachkou;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The article is an essay on the development of technologies for natural language processing, which formed the basis of BERT (Bidirectional Encoder Representations from Transformers), a language model from Google, showing high results on the whole class of problems associated with the understanding of natural language. Two key ideas implemented in BERT are knowledge transfer and attention mechanism. The model is designed to solve two problems on a large unlabeled data set and can reuse the identified language patterns for effective learning for a specific text processing problem. Architecture Transformer is based on the attention mechanism, i.e. it involves evaluation of relationships between input data tokens. In addition, the article notes strengths and weaknesses of BERT and the directions for further model improvement.

  • Open Access Russian
    Authors: 
    Rykhard Bohush; I. Yu. Zakharava; Sergey Ablameyko;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    In the paper the algorithm for object detection in high resolution images is proposed. The approach uses multiscale image representation followed by block processing with the overlapping value. For each block the object detection with convolutional neural network was performed. Number of pyramid layers is limited by the Convolutional Neural Network layer size and input image resolution. Overlapping blocks splitting to improve the classification and detection accuracy is performed on each layer of pyramid except the highest one. Detected areas are merged into one if they have high overlapping value and the same class. Experimental results for the algorithm are presented in the paper.

  • Open Access Russian
    Authors: 
    Yu. N. Sotskov; Natalja G. Egorova; Natalja M. Matsveichuk;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    Time-management refers to planning the available time in correspondence with personal goals and preferences of the employee. The criterion for efficient time-management lies in achieving definite goals in minimum possible time. This paper presents a short review of the literature on time-management. It is shown how scheduling theory may be used for optimal planning in time-management. The problem of minimizing the total (average) weighted completion time of the planned jobs by an employee is considered provided that only lower and upper bounds of the possible processing time of each job are known before scheduling. Software for constructing a permutation of the given jobs with the largest relative semi-perimeter of the optimality parallelepiped has been developed. Computational experiment on the computer showed the effectiveness of the developed algorithms for constructing schedules for time-management.

  • Open Access Russian
    Authors: 
    M. L. Radziukevich; V. F. Golikov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The main options for the formation of a shared secret using synchronized artificial neural networks and possible patterns of behavior of a cryptanalyst are considered. To solve the problem of increasing the confidentiality of the generated shared secret, if it is used as a cryptographic key, it is proposed to use the mixing a certain number of results of individual synchronizations (convolution). As a mixing function, we consider the convolution of the vectors of network weights by bitwise addition modulo 2 of all the results of individual synchronizations. It is shown that the probability of success of a cryptanalyst is reduced exponentially with an increase of the number of terms in the convolution and can be chosen arbitrarily small. Moreover, the distribution law of the generated key after convolution is close to uniform and the uniformity increases with the number of terms in the convolution.

  • Open Access Russian
    Authors: 
    A. Y. Kharin;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    In the problems of data flows analysis, the problems of statistical decision making on parameters of observed data flows are important. For their solution it is proposed to use sequential statistical decision rules. The rules are constructed for three models of observation flows: sequence of independent homogeneous observations; sequence of observations forming a time series with a trend; sequence of dependent observations forming a homogeneous Markov chain. For each case the situation is considered, where the model describes the observed stochastic data with a distortion. "Outliers" ("contamination") are used as the admissible distortions that adequately describe the majority of situations appear in practice. For such situations the families of sequential decision rules are proposed, and robust decision rules are constructed that allow to reduce influence of distortion to the efficiency characteristics. The results of computer experiments are given to illustrate the constructed decision rules.

  • Open Access Russian
    Authors: 
    N. N. Guschinsky; O. Battaia; A. Dolgui;
    Publisher: HAL CCSD
    Country: France

    A problem of design of a flow line composed of stationary machines for sequential batch machining is considered. A MIP model for synthesis of manufacturing process structure is proposed. The model takes into account main technological and structural constraints with the criterion of the minimal line equipment cost. Results of solving a real industrial example are also presented.

  • Open Access Russian
    Authors: 
    Y. A. Drybin; S. V. Sadau; V. S. Sadau;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    It is shown that the choice of the time sampling parameter of the digital model of a continuous dynamic system with chaotic modes based on its dynamics makes it possible to control the characteristics of the output sequence, including avoiding short cycles and periodic behavior modes. On the example of the Lorentz system, the analysis of the law of motion of a chaotic system, linearized in the vicinity of points of stable and unstable equilibrium, is carried out, on the basis of which the parameters of the mathematical model of the generator of pseudo-random numbers are selected. The output sequence of numbers generated in proposed way is subjected to statistical and correlation analysis. Based on the results of the tests carried out, we can say that the obtained pseudo-random sequences based on continuous chaotic systems have statistically random properties and can be used in steganographic and cryptographic systems.

  • Open Access Russian
    Authors: 
    Yu. V. Pottosin;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The orthogonal disjunctive normal forms (DNFs) of Boolean functions have wide applications in the logical design of discrete devices. The problem of DNF orthogonalization is to get for a given function such a DNF that any two its terms would be orthogonal, i. e. the conjunction of them would be equal identically to zero. An approach to solve the problem using the means of graph theory is suggested. The approach is proposed by representation of the function as perfect DNF. Obtaining all the intervals of the Boolean space where the given function has value 1 is supposed, and the intersection graph of those intervals is considered. Two methods to obtain a minimum orthogonal DNF are considered. One of them reduces the problem toward finding out the smallest dominating set in the graph by covering its vertices with their closed neighborhoods, the other - to obtain the maximum independent set by lexicographic enumeration. It is shown how the suggested approach can be extended on incompletely specified Boolean functions.

  • Open Access Russian
    Authors: 
    R. S. Vashkevich; E. S. Azarov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    The paper investigates the problem of voice activity detection from a noisy sound signal. An extremely compact convolutional neural network is proposed. The model has only 385 trainable parameters. Proposed model doesn’t require a lot of computational resources that allows to use it as part of the “internet of things” concept for compact low power devices. At the same time the model provides state of the art results in voice activity detection in terms of detection accuracy. The properties of the model are achieved by using a special convolutional layer that considers the harmonic structure of vocal speech. This layer also eliminates redundancy of the model because it has invariance to changes of fundamental frequency. The model performance is evaluated in various noise conditions with different signal-to-noise ratios. The results show that the proposed model provides higher accuracy compared to voice activity detection model from the WebRTC framework by Google.

  • Open Access Russian
    Authors: 
    Vladimir Sapozhnikov; Vl.V. Sapozhnikov; D. V. Efanov;
    Publisher: The United Institute of Informatics Problems of the National Academy of Sciences of Belarus

    Simpler than known structure of the system with error correction in calculations is proposed based on duplication and triplication of blocks with majority principle of choosing the values of signals. It is advisable to use the new fault-tolerant structure for automation devices with combinational logic. In fault-tolerant structure synthesis, the parity method is used to establish the fact of a fault in the main logic unit and the logical complement method is used determine incorrectly calculated output functions and to generate signals for their correction. The method also allows to adjust the values of incorrectly calculated functions. Structural diagram and description of error correction system are given. The synthesis algorithm of control equipment is described with minimization of the technical implementation complexity. The experiment results with control combinational circuits are given, confirming the high efficiency of proposed system structure with error correction.

Send a message
How can we help?
We usually respond in a few hours.