CC BY
7Extended methodology for deriving formal concepts
V. A. Semenova, S. V. Smirnov
Samara Federal Research Scientific Center RAS, Institute for the Control of Complex Systems RAS
Email: smirnov@iccs.ru
DOI 10.24412/cl-35065-2021-1-02-73
In the study we analyze two methodologies for deriving formal concepts: the classical one, which focuses
on the posterior analysis of the object�s properties of the studied knowledge domain [1], and non-classical, the
cornerstone of which is the priori formation of the set of measured object�s properties and the determination
of existential relations on this set [2]. Firstly, in the technological chain of the target transformation of the
source data we fix a position where the difference between considered methodologies really shows itself. Sec-
ondly, we establish the commonality of these two approaches in the aspect of the unity of their hypothetical-
deductive basis. We demonstrate the need for the joint use of the considered methodologies at processing
incomplete and inconsistent empirical data about studied knowledge domain [3].
This work was supported by Ministry of Science and Higher Education of the Russian Federation, R&D registration
numbers AAAA-A19-119030190053-2.
References
1. Ferre S., Huchard M., Kaytoue M., Kuznetsov S.O., Napoli A. Formal Concept Analysis: From Knowledge Discovery
to Knowledge Processing // A Guided Tour of Artificial Intelligence Research. V. II. AI Algorithms. Springer Int. Publishing.
2020. P. 411-445.
2. Lammari N., du Mouza C., Metais. E. POEM: an Ontology Manager based on Existence Constraints // Int. Conf.
DEXA, 2008. Lect. Notes in Computer Science, V. 5181. Springer-Verlag Berlin Heidelberg. P. 81-88.
3. Samoylov D.E., Semenova V.A., Smirnov S.V. Defuzzification of the initial context in Formal Concept Analysis // Int.
Conf. on Information Technology and Nanotechnology, 2019. CEUR Workshop Proc. V. 2416. P. 1-9.
Pattern recognition for bubbly flows with vapor or gas-liquid interfaces using U-Net architecture
A. V. Seredkin1,2, I. P. Malakhov1,2, V. S. Serdyukov1,2, R. I. Mullyadzhanov1,2, A. S. Surtaev1,2
1Novosibirsk State University
2Institute of Thermophysics SB RAS
Email: a.seredkin@g.nsu.ru
DOI 10.24412/cl-35065-2021-1-02-74
We apply deep learning algorithms to images of the water pool boiling in order to detect the evolution of
the vapor bubbles. It can allow us to simultaneously analyze each separated bubble and obtain all key parame-
ters including the bubble growth rate, departure diameter and time and nucleation site�s activation. For analy-
sis the data obtained in experiments at pool boiling conditions in pressure range 9-101 kPa were taken for the
basis [1]. The camera is directed upwards and focuses on the heat substrate itself, which makes it easier for
the network to distinguish surface bubbles from one that floats up. Using our technique the evolution of each
bubble was automatically measured based on video recording of the experiment. As a basic network U-Net
with ResNet 50 encoder was used [2]. The network was trained on manually labeled dataset augmented with
random rotations, flips, extra surface bubbles, background bubbles and background noise. We demonstrated
the capabilities by tracking the appearance, growth and separation of the bubbles from the heating surface,
which can be used for the mechanistic analysis of the heat transfer rate during liquid boiling at various pres-
sures.
This work was supported by RFBR and TUBITAK according to the research project � 20-58-46008.
References
1. Surtaev, A., Serdyukov, V., & Malakhov, I. (2020). Effect of subatmospheric pressures on heat transfer, vapor
bubbles and dry spots evolution during water boiling. Experimental Thermal and Fluid Science, 112, 109974.
2. Python library with Neural Networks for Image Segmentation [Electron. resource]. URL: https://github.com/
qubvel/segmentation_models (the date of access: 01.06.2021).
Approach to automatic population of ontologies of scientific subject domain using lexico-syntactic patterns
Y. A. Zagorulko, E. A. Sidorova, I. R. Akhmadeeva, A. S. Sery
A. P. Ershov Institute of Informatics System, SBRAS
Email: zagor@iis.nsk.su
DOI 10.24412/cl-35065-2021-1-02-75
The paper presents an approach to automatic population of ontologies of scientific subject domain (SSD)
using lexico-syntactic patterns and corpus of texts related to modeled domain. These patterns are built on the
basis of ontology design patterns (documented descriptions of practical solutions to typical problems of onto-
logical modeling) [1] provided by the system for the automated development of SSD ontologies [2].
This research was supported by the Russian Foundation for Basic Research (grants No. No. 19-07-00762).
References
1. Gangemi A., Presutti V. Ontology Design Patterns // Handbook on Ontologies, Eds., Staab, S. and R. Studer. Berlin:
Springer Ver-lag, 2009. P. 221-243.
2. Zagorulko Yu. A., Borovikova O. I. Using a System of Heterogeneous Ontology Design Patterns to Develop
Ontologies of Scientific Subject Domains // Programming and Computer Software, 2020, V. 46, No. 4, P. 273�280.
