|Article title||HIGH-EFFICIENCY METHOD AND ALGORITHMS OF AUTOMATIC CLASSIFICATION OF OBJECTS IN THE CONDITIONS OF PARAMETRICAL UNCERTAINTY AND CLASS INTERSECTION ON THE BASISOF METHODOLOGYWITH SYSTEM MAXIMIZATION OF ENTROPY|
|Section||SECTION I. DATA ANALYSIS AND KNOWLEDGE MANAGEMENT|
|Month, Year||07, 2016 @en|
|Abstract||In the report the problem of objects classification in the conditions of parametrical uncer-tainty and class intersection is considered. Specificity of the problem is absence of reliable infor-mation about parameters of a priori known classes presented in the catalogue of reference values in the form of confidence intervals of signs. Such kind of uncertainty imposes essential restrictions on a range of methods and algorithms which can be used for the solution of the problem. In the report G.V. Sheleykhovskiy"s method of objects classification, which takes into account required restrictions and possessing high degree of reliability of classification is considered. The basic imperfections of the method complicating its application in real conditions are shown. On the basis of G.V. Sheleykhovskiy"s method the combined method of the classification possessing improved productivity and devoid of known imperfections of the basic method is developed. The combined method assumes improvement of the base G.V. Sheleykhovskiy"s method by the relative positioning analysis in parametrical space of the classification objects presented in the form of points and a priori known classes from the catalogue of reference values, representing areas in the parametrical space. In the document the following algorithms realising the developed method are presented: algorithm of the entrance data preliminary analysis, algorithm of the classification matrix reduction, algorithm of list representation of the classification matrix, algorithm of classification results analysis and formation of information data of new objects. The developed method allows to reduce considerably computing labouriousness of a base method, and also to relieve it of a number of essential defects. So, for example, the known problem of G. V. Sheleykhovskiy’s method convergence is solved. In article existence of specific objects which influence process of convergence of this method is proved. The offered way of a solution of a convergence problem of a base method of classification differs from known ways that it does not demand increase in expenses of computing and time resources. In work the performed experimental researches are described and the received results are presented. The present results allow to draw a conclusion on decrease in the general of computing labouriousness at classification of objects with use of the combined method in relation to the method of classification of G.V. Sheleykhovskiy on the average twice.|
|Keywords||Genetic algorithm; clustering; information retrieval; prediction.|
|References||1. Gonsales R., Tu Dzh. Printsipy raspoznavaniya obrazov [Principles of pattern recognition]. Moscow: Mir, 1978, 411 p.
2. Simankov V.S., Lutsenko E.V. Adaptivnoe upravlenie slozhnymi sistemami na osnove teorii raspoznavaniya obrazov [Adaptive control of complex systems based on the theory of pattern recognition]. Krasnodar: Tekhn. un-t Kuban. gos. tekhnol. un-ta, 1999, 318 p.
3. Kirillov N.P. Priznaki klassa i opredelenie ponyatiya «tekhnicheskie sistemy» [Class character-istics and the definition of "technical systems"], Aviakosmicheskoe priborostroenie [Aerospace Instrumentation], 2009, Issue 8, pp. 32-38.
4. Zhuravlev Yu.I. Ob algebraicheskom podkhode k resheniyu zadach raspoznavaniya i klassi-fikatsii [About algebraic approach to solving the problems of recognition and classification], Problemy kibernetiki [Problems of Cybernetics], 1978, Issue 33, pp. 5-68.
5. Dorogov A.Yu. Teoriya i proektirovanie bystrykh perestraivaemykh preobrazovateley i slabosvyazannykh neyronnykh setey [Theory and design fast tunable transmitters and loosely-coupled neural networks]. St. Petersburg: Politekhnika, 2014, 344 p.
6. Terano T., Asai K., Sugeno M. Prikladnye nechetkie sistemy [Applied fuzzy systems]. Moscow: Mir, 1993, 368 p.
7. Kruchinin I.I., Loginov B.M. Modelirovanie protsessov klassifikatsii mnogomernykh ob"ektov peresekayushchikhsya klassov na osnove predstavleniy nechetkikh mnozhestv [Modeling of processes of classification of multidimensional objects overlapping classes based on the concepts of fuzzy sets], Trudy 2-y Rossiyskoy konferentsii molodykh uchenykh po matematicheskomu modelirovaniyu [Proceedings of 2-nd Russian conference of young scientists on mathematical modeling]. Moscow: MGTU, 2002, pp. 14-23.
8. Voloshin G.Ya. Metody raspoznavaniya obrazov konspekt lektsiy. Sayt tsifrovykh uchebno-metodicheskikh materialov VGUES [Pattern recognition lecture notes. The website of digital educational and methodological materials VSUES]. Available at: http://abc.vvsu.ru/Books/Metody_r.
9. Zhuravlev Yu.I. Raspoznavanie obrazov i raspoznavanie izobrazheniy. Raspoznavanie, klassi-fikatsiya, prognoz [Pattern recognition and image recognition. Recognition, classification, pre-diction], Matematicheskie metody i ikh primenenie [Mathematical Methods and their Applica-tion], 1989, Issue 2, pp. 5-73.
10. Vapnik V.N., Chervonenkis A.Ya. Teoriya raspoznavaniya obrazov [Theory of pattern recogni-tion]. Moscow: Nauka. 1974, 416 p.
11. Kruchinin I.I., Loginov B.M. Modelirovanie protsessov klassifikatsii mnogomernykh ob"ektov peresekayushchikhsya klassov na osnove gibridnykh neyrosetey [Modeling of processes of classification of multidimensional objects overlapping classes based on hybrid neural networks], Trudy 2-y Rossiyskoy konferentsii molodykh uchenykh po matematicheskomu modelirovaniyu [Proceedings of 2-nd Russian conference of young scientists on mathematical modeling]. Moscow: MGTU, 2002, pp. 24-30.
12. Krug P.G. Neyronnye seti i neyrokomp'yutery: uchebnoe posobie po kursu «Mikroprotsessory» [Neural networks and Neurocomputers: textbook for the course "Microprocessors"]. Moscow: Izd-vo MEI, 2002, 176 p.
13. Bregman L.M. Dokazatel'stvo skhodimosti metoda Sheleykhovskogo dlya zadachi s transportnymi ogranicheniyami [The proof of convergence of the method Celakovskeho for the problem with transport constraints], Vychislitel'naya matematika i matematicheskaya fizika [Computational mathematics and mathematical physics]. Moscow, 1967, pp. 147-156.
14. Grishkov A.F., Kulakov A.A., Shpak V.F. Klassifikatsiya obrazov radiotekhnicheskikh signalov [Classification of images of radio signals], Voprosy spetsial'noy radioelektroniki [Questions of Special Radio Electronics], 2009, Issue 2, pp. 30-37.
15. Trukhaev R.I. Modeli prinyatiya resheniy v usloviyakh neopredelennostey [Models of decision making under conditions of uncertainty]. Moscow: Nauka, 1981, 258 p.
16. Jaynes, E.T. Information Theory and Statistical Mechanics, Physical Review, 1957, No. 106 (4), pp. 620-630. DOI:10.1103/PhysRev.106.620. Bibcode: 1957PhRv..106..620J.
17. Jaynes E.T. Information Theory and Statistical Mechanics II, Physical Review, 1957, No. 108 (2), pp. 171-190. DOI: 10.1103/PhysRev.108.171. Bibcode: 1957PhRv..108..171J.
18. Shpak B.F., Getmanchuk A.V., Radchenko S.A., Kulakov A.A. K voprosu o skhodimosti algo-ritma klassifikatsii radiotekhnicheskikh obrazov po metodu G.V. Sheleykhovskogo [To the question of convergence of the algorithm for the classification of electronic images by the method of G. V. Celakovskeho], Voprosy spetsial'noy radioelektroniki [Questions of Special Radio Electronics], 2011, Issue 2, pp. 41-46.
19. Sinkhorn, Richard. A relationship between arbitrary positive matrices and doubly stochastic matrices. Ann. Math. Statist, 1964.
20. Sinkhorn, Richard, & Knopp, Paul. Concerning nonnegative matrices and doubly stochastic matrices. Pacific J. Math, 1967.