V. N. Pavlysh, L. A. Lazebnaya.
THE MATHEMATICAL MODELS ANDCONTROL ALGORITHMS OF DYNAMIC TREATMENT PROCESSESON
ANISOTROPIC UNDERGROUND MASSIFS.
UDC 681.518.52:622.53 Annotation: The results of determined mathematical models construction and automatic control algorithms of
dynamic treatment processes on coal stratum as anisotropic underground massif for its dangerous properties reducing during underground mining are represented. Three main treatment technological schemes are considered, the base of algorithms is requirement of active forces uniform distribution in deal area, which attracts the physics of process. Keywords: process, mathematical model, control, algorithm, treatment.
V. N. Pavlysh, G. B. Peretolchina.
THE BASE OF MATHEMATICAL MODEL OF LIQUID
MOTION PROCESS IN ANISOTROPIC COAL STRATUM.
UDC 622.734.001.57 Annotation: In the article the task of construction and investigation of determine mathematical model of liquid motion processes in anisotropic coal stratum during its hydraulic action for dangerous properties reducing is considered. The model based on particular derivatives equations, describes nonlinearelastic filtration regime of compress-resist liquid in strength continuous environment, filtration parameters are given as random quantities with fixed variation interval. Computer realization made by ending-difference approximation. Keywords: the mathematical model, process, parameter, equation, algorithm.
S. V. Bespalova, S. M. Romanchuk, T. V. Yermolenko, V. I. Bondarenko.
CONSTRUCTION OF PREDICTIVE MODELS OF THE PARAMETERS
OF WATER PRESSURE IN WATER DISTRIBUTION NETWORKS BY MEANS
OF MACHINE TRAINING METHODS.
UDC 004.942 Annotation: The effectiveness of machine learning methods for predicting the behavior of the water supply
network is searched in the article. Predictive models of water pressure values at control points on the network are constructed. A comparative analysis of the quality of the constructed models is made. Keywords: multicollinearity, gradient boosting, neuronet.
M. V. Blizno.
STUDY OF REGULARITIES OF DEVELOPMENT OF THE NON-VOLATILE MEMORY.
UDC 622.734.001.57 Annotation: The article presented informational material for non-volatile memory used in the computer industry, as well as regularities of its development. Keywords: ROM, Flash, nonvolatile memory, flash memory.
O. O. Varlamov.
ON CREATION ON THE BASIS OF MIVAR DECISIONMAKING SYSTEMS «ROBO!RAZUM» AUTONOMOUS HARVESTERS AND TRACTORS
FOR AGRICULTURE.
UDC 004.82+007.52 Annotation: In the field of robotics, «ROBO! MIVAR» decision-making systems can be used to control autonomous combines and tractors that, in the process of moving across the field, can additionally perform a variety of activities, as well as work in group interaction mode, in different weather conditions and even with faulty technical vision. Keywords: mivar, mivar networks, artificial intelligence, unmanned vehicles, decision-making
systems, KESMI, robotic systems, cyber-physical systems, «ROBO!MINUM», autonomous combines and tractors for agriculture.
A. V. Nicenko, V. Ju. Shelepov, S. A. Bolshakova.
ON THE DEPENDENCY TREE FOR A SIMPLE EXTENDED RUSSIAN SENTENCE.
UDC 004.89:004.93 Annotation: The paper deals with the the principle of automatically creating a dependency tree for the Russian language sentences. The problem is solved on the basis of identifying the word forms that constitute the sentence by searching them in the morphological dictionary. A feature of the proposed method is the use of a set of rules for the choice of homonyms and the definition of dependencies between the words of a sentence. The article provides a set of data rules and examples of their work. As a result, a depending word or an empty string for main word is connected to each word of the sentence. Graphically, this is displayed as a tree structure, reflecting the dependencies between the words of a sentence. Keywords: automatic text analysis, dependency tree, homonyms, morphological analysis, syntactic analysis, dependency rules.
Ya. S. Pikalyov.
THE DEVELOPMENT OF THE AUTOMATIC TRANSFORMATION OF ENGLISH ACCENTS IN RUSSIAN
TEXTS WITH THE APPLICATION OF DEEP LEARNING.
UDC 004.912 Annotation: This work is devoted to the task of developing an automatic system for transforming English inserts in Russian texts using deep learning. The author proposed a hybrid method of transcription, which was developed based on the work of linguists, as well as using deep learning. This paper presents a declarative-procedural approach, using both the dictionary and the rules of English-Russian practical transcription, for the transformation of English inserts found in Russian texts. The dictionary of EnglishRussian practical transcription using the finite state machine mechanism is prepared. Trained neural network for the classification of the language of the text using a multilayer convolutional neural
networks. A neural network for the transformation of words in Latin not found in the dictionary using neural network architecture such as encoder-decoder has been trained. Keywords: natural language processing; practical transcription; finite state machine
mechanism; transformer; convolutional neural networks.
V. J. Przekop.
MATHEMATICAL MODELS OF THE GROWTH RATES OF THE FINANCIAL INSTRUMENTS ON THE BASIS OF SYMMETRIC AND ASYMMETRICAL
LAPLACE DISTRIBUTION.
UDC 622.734.001.57 Annotation: In the paper it is proposed mathematical models of the process of growth of stock indicators on the basis of symmetric (two-parameter model) and asymmetric (three-parameter model) Laplace
distribution using gamma distribution. Keywords: Laplace distribution, the gamma distribution, random variable,
mathematical model.
A. A. Kharlamov, D. I. Gordeev.
DISTRIBUTIVE VS NETWORK SEMANTICS IN DIALOG SYSTEMS.
UDC 004.942, 514.18, 519.652 Annotation: In the last 8 years, increased interest in the field of dialogue agents has resumed. This is largely due to the introduction of machine learning in the tasks of automatic processing of the natural language. The use of means of distributive and network semantics allows using generalized data from huge corpus of texts, which was more problematic when using n-grams. Also, new language models, trained in huge buildings, can significantly reduce the cost of additional training for new tasks (transfer learning), and in some cases even do without it (zero-shot learning). In addition, this chapter examines the established neural network architectures and promising approaches to the use of neural networks for the tasks of
automatic processing of the language in general, and interactive agents in particular. Also provides an overview of the modular approach to interactive agents. The main types of modules are considered. Keywords: interactive agents; NLP; natural language processing; neural networks; deep learning; distributive semantics; conversational agents; Word2Vec; Elmo; Bert.
РУС
ENG
