A. A. Kharlamov.
PROSPECTS FOR THE DEVELOPMENT
OF SEMANTIC REPRESENTATIONS BASED
ON TRENDS IN NEUROINFORMATICS.
UDC 004.942, 514.18, 519.652 Annotation: A unified approach to the processing of textual and visual information is rationalized;this makes it
possible to talk about the construction of effective multimodal systems for information processing
and representation, which is the only correct approach to the evolutionin modelsof human
intellectual functions. One of the capabilities of informational combination to represent knowledge
bases – the core of modern information systems – is the use of semantic networks, since it is in
this form that knowledge is presented in human consciousness. An example of such an
implementation is the TextAnalyst software technology for automatic semantic analysis of
unstructured texts, which effectively implements the formation of homogeneous semantic networks,
as well as text classification and clustering, etc. Keywords: integration of information technology, semantic networks as the basis for
standardizing the representation of knowledge, semantic networks in the human
consciousness, semantic analysis of texts, integration of modalities.
S. S. Antsyferov, K. N. Fazilova. METHODOLOGY OF COGNITIVE SYSTEMS NON-EQUILIBRIUM STABILITY EVALUATING.
UDC 681.518.9; 621.384.3 Annotation: Cognitive systems, as self-organizing systems, have a number of special properties, including the
fundamental non-equilibrium of the state, which provides the possibility of transformation of the
structure. As an adequate model of non-equilibrium stability of a cognitive system, we propose to
use a nonlinear equation based on a probabilistic representation of the effectiveness of the
functioning of interrelated structural elements of the system. Numerical modeling allowed us to
determine the dependence of the system state on its structural construction, the number of
structural elements and their efficiency. Based on the use of the nonequilibrium stability model, the
conditions for nonequilibrium stability were determined. Keywords: cognitive system, structural element, nonequilibrium stability,
structure transformation, structural elements efficiency.
R. A. Bushuev, A. V. Devyatkin, D. A. Zykov, V. M. Shoskalne. APPLICATION OF MIVAR DATABASES AND SITE RULES FOR PEOPLE WITH VISUAL DISABILITIES.
UDC 004.021 Annotation: The article is devoted to the problem of rehabilitation of disabled people, including the interaction of
visually impaired and blind people with computers. Existing solutions, called assistive technologies,
can partially eliminate the social barrier of the information space. The article reflects the creation of
an accessible software environment, the use of which will not cause misunderstanding for people
with or without vision problems. In this regard, it was decided the use of CASMI designer of expert
systems mivar. A tool for creating knowledge models with an unlimited number of relationships,
parameters, and relationships, with logical output. Keywords: mivar databases, AudCat, mivar, mivar networks, Reasoner, Wi!Mi.
S. A. Zori, O. A. Krivodubsky. REPRESENTATION OF A MULTI-LEVEL ENTERPRISE MANAGEMENT SYSTEM AS AN ACTIVE NEURAL NETWORK.
UDC 590.1 Annotation: The paper considers a technique that combines elements of the classical representation of control
systems and components of intelligent techniques in the form of an active neural network and
image recognition. Thus, for each level of management of production units are considered active
centers of the neural network, with the inherent mathematical characteristics, upon which to build
algorithms, subsystems and control, parametric identification of models, as an integral part of this
active learning phase of the neural network. Keywords: mathematical models, images, mapping, equivalence, parametric identification.
V. V. Klimov, M. N. Cheptsov. BASE STATION CONTROLLER BUFFER’S SIZE ESTIMATING PREDICTION METHOD.
UDC 004.005 Annotation: The article discusses the problem of predicting base station controller's buffer size estimate.
A discrete-event model of the scheduler operation in the Max-Plus algebra basis, taking into
account the query processing delays, is presented. It allows to display the dynamics of the planner
and improves the accuracy of forecasting. Keywords: buffer size estimation, Max-Plus algebra, the query processing delays.
A. S. Minenko, T. V. Vanzha. HUMAN EMOTIONAL RECOGNITION SYSTEM.
UDC 004.93 Annotation: The main human emotions are highlighted in the work, a brief review of the history of the
emergence of emotions and emotional state is carried out. In the work, an algorithm and a system
for recognizing emotions to determine the emotional state are developed. The main stages of
automatic recognition of emotions are highlighted. The process of detecting a face by the Viola-Jones
method and its description are considered. The characteristic of emotions is given by the
combination of eyebrow and mouth points. Keywords: emotional state, human-machine interaction, key points,
recognition of emotions.
M. L. Poviraeva. APPLICATION OF MIVAR DATABASES AND RULES FOR THE SUBJECT AREA «MONITORING, MAINTENANCE AND REPAIR OF GROUND-BASED DEEP SPACE MANAGEMENT SYSTEMS».
UDC 004.021 Annotation: The article is devoted to the issue of issuing recommendations to a repairman on the basis of indicators
of sensors installed on the equipment. The article discusses existing monitoring systems. A solution is
proposed to give recommendations to the repairman on the basis of indicators of sensors installed on
the equipment using mivar databases to reduce the time spent by the repairman on identifying the
problem link among all the equipment installed at the facility. The proposed solution was analyzed with
the advantages and disadvantages. The conclusion is drawn on the practicality and feasibility of
applying mivar technologies in the field of monitoring, maintenance and repair of ground-based deep
space management systems. Keywords: Monitoring, maintenance, repair, ground-based deep space management
systems, mivar, mivar networks, Reasoner, Wi!Mi.
I. S. Salnikov, R. I. Salnikov, T. D. Klyushanova. SYSTEM FOR COMPUTER DIAGNOSTICS OF HUMAN PSYCHOPHYSIOLOGICAL CONDITIONS.
UDC UDC 159.923:159.942.5 Annotation: The article provides information on the system of complex computer diagnostics of psychophysiological
states of a person developed at the Institute and examines the currently promising,
but not yet formalized for computer use, methods, means and methods of drug-free intellectual and
spiritual rehabilitation and self-regulation of psychoemotional states of a person, provides a brief
biographical information about the life and work of well-known supporters or adherents of these
methods, means and methods of drug-free therapy. Keywords: computer diagnostics of psychophysiological states of a person, methods,
methods and means of drug-free intellectual and spiritual rehabilitation and self-regulation
of psychoemotional states of a person, adherents of drug-free therapy.
O. V. Rychka. EFFICIENCY ANALYSIS OF THE IMPROVED METHODS OF SEARCHING AND PROCESSING ANOMALIES FOR NONLINEAR MODELS WITH INTERNAL LINEAR.
UDC 519.25 Annotation: The article deals with the problem of detecting and processing anomalous data in order to improve the
quality of predictive regression models. The main disadvantages of existing methods for detecting
unreliable data are revealed. Improved methods for finding and correcting anomalous data are
proposed. The main criteria for evaluating the effectiveness of the developed methods are described.
Numerical studies of the effectiveness of the methods for nonlinear models with internal linearity are
performed. Keywords: anomalous data, regression analysis, nonlinear regression, linear model,
performance criteria.
Ya. S. Pikalyov, T. V. Yermolenko. ADAPTATION OF ALBERT NEURAL NETWORK MODEL FOR LANGUAGE MODELING PROBLEM.
UDC 004.912 Annotation: The article discusses the main features of the ALBERT deep network architecture, its advantages
and disadvantages for use in the task of language modeling. A modification of the network for
applying the two-stage training procedure, using the attention mask for the encoder during
preliminary training, is proposed. Studies have shown that an increase in the number of hidden
layers and the size of a hidden layer leads to an increase in the accuracy of the model, and
positional encoding is an unnecessary operation. Using AdaptiveSoftmax does not affect the word
error and will improve network performance. Keywords: language model, Transformer architecture, BERT, ALBERT, text masking,
BPE-tokenization.
РУС
ENG
