350 rub
Journal Neurocomputers №3 for 2018 г.
Article in number:
Atomic machine learning
Type of article: scientific article
UDC: 519.67, 519.23:24, 517.5
Authors:

S.Yu. Eremenko − Dr.Sc. (Eng.), Professor, Director of Soliton Scientific Pty Ltd (Sydney, Australia) 

E-mail: sergei@solitonscientific.com

Abstract:

It is shown how the theory of Atomic functions known since 1970th  and their AString generalizations can be expanded to Machine Learning algorithms to represent atomic kernels in Support Vector Machines, Density Estimation, Principal Component Analysis, LOESS regression, and be the foundation of atomic regression, atomic activation function for Neural Networks and atomic computer. Representing a probabilistic weighted uniform distribution, these functions are related to Prouhet–Thue–Morse ‘fair game’ sequence and offer a fast calculation of derivatives. Multiple Matlab, R and Python examples calculate complex regressions, decision boundaries and precise target clusters important for medical, military and engineering applications. In combination, these methods compose Atomic Machine Learning theory which extends the 50 years’ history of atomic functions and their generalisations to new scientific domains.

Pages: 13-26
References
  1. Eremenko S.Yu. Atomic Strings and Fabric of Spacetime. Foreign Radioelectronics: Achievements of modern radioelectronics. 2018, Accepted for publication. https://www.researchgate.net/publication/320457739.
  2. Rvachev V.L., Rvachev V.A. About one Finite Function, DAN URSR. A (6). 705–707. 1971.
  3. Rvachev V.L., Rvachev V.A. Non-classical methods in the theory of approximations in boundary value problems. Kiev: Naukova Dumka. 1982. 196 p.
  4. Rvachev V.L. Theory of R-functions and their applications. Kiev: Naukova Dumka. 1982.
  5. Kravchenko V.F., Rvachev V.A. Application of atomic functions for solutions of boundary value problems in mathematical physics. Foreign Radioelectronics // Achievements of modern radioelectronics. 1996. № 8. P. 6–22.
  6. Kravchenko V.F., Kravchenko O.V., Pustovoit V.I, Churikov D.I. Atomic, WA-Systems, and R-Functions Applied in Modern Radio Physics Problems: Part 1. Journal of Communications Technology and Electronics. 2014, V. 59. №10. P. 981– 1009. https://www.researchgate.net/publication/298406834. DOI: 10.1134/S1064226914090046.
  7. Kravchenko V.F., Kravchenko O.V., Pustovoit V.I, Pavlikov V.V. Atomic Functions Theory: 45 Years Behind. https://www.researchgate.net/publication/308749839. DOI 10.1109/MSMW.2016.7538216.
  8. Kravchenko V.F. Lectures on the theory of atomic functions and their applications. Moscow: Radiotechnika. 2003.
  9. Kravchenko V.F., Rvachev V.L. Logic Algebra, atomic functions and wavelets in physical applications. Moscow: Fizmatlit. 2009.
  10. Eremenko S.Yu. Finite Element Methods in Mechanics of Deformable Bodies. Kharkov: Osnova. 1991. 272p. https://www.researchgate.net/publication/321171685. 
  11. Eremenko S.Yu. Natural Vibrations and Dynamics of Composite Materials and Constructions. Kiev: Naukova Dumka. 1992. 182 p. 
  12. Eremenko S.Yu., Rvachev V.L. A survey of variants of the structural finite-element method // Journal of Mathematical Sciences. 1998. V. 90. № 2. P. 1917–1922.
  13. Eremenko S.Yu. Combined Non-Archimedean Calculi. Foreign Radioelectronics // Achievements of modern radioelectronics. 1996. № 8. P. 57–65. https://www.researchgate.net/publication/320616255. 
  14. Eremenko S.Yu., Kravchenko V.F., Rvachev V.L. Combined Non-Archimedean Calculi and New Models of Relativistic Mechanics. Foreign Radioelectronics // Achievements of modern radioelectronics. 1997. № 9. P. 26–38. https://www.researchgate.net/publication/320616269.
  15. 15. Gotovac H., Kozulic V, Gotovac B. Space-Time Adaptive Fup Multi-Resolution Approach for Boundary-Initial Value Problems // Computers, Materials and Continua. 2010. 15(3). P. 173–198. https://www.researchgate.net/publication/287948119 16. Sirodzha I.B. Quantum model and methods of artificial intelligence for solution taking and control. Kiev: Naukova Dumka. 2002. 490 p. 

  16. Kolodyazhni V.M., Rvachev V.A. Atomic functions. Generalizations to the case of several variables and perspective directions of practical application // Cybernetics and Systems Analysis. 2007. V. 43. № 6. P. 155–177.
  17. Fabius J. A probabilistic example of a nowhere analytic C -function. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete, 1966, 5 (2): 173–174, doi:10.1007/bf00536652, MR 0197656.
  18. Hilberg W. Impulse und Impulsfolgen, die durch Integration oder Differentiation in einem veranderten Zeitmastab reproduziert warden. AEU. (25). 1. 1971. P. 39–48.
  19. Palacios-Huerta I. Tournaments, fairness and the Prouhet–Thue–Morse sequence. Economic inquiry, 2012. 50 (3):  848–849. DOI:10.1111/j.1465-7295.2011.00435.x.
  20. Kravchenko V.F., Kravchenko O.V., Konovalov Ya.Yu. Prouhet-Thue-Morse sequence and atomic functions in applications of physics and techniques. Journal of Measurement Science and Instrumentation, 2015, 6(2): 128–141.
  21. https://www.researchgate.net/publication/280316794. DOI: 10.3969/j.issn.16748042.2015.02.005.

  22. Filippov A.T. The Versatile Soliton. Modern Birkhauser Classics. 2000. 261 p.
  23. Bougaev A.A., Urmanov A.M., Gross A.M., Tsoukalas L.H. Method of Key Vectors Extraction Using R-Cloud Classifiers. Springer, 2006. DOI 10.1007/1-4020-5263-4_15. 
  24. Raschka S. Python in Machine Learning. PACKT Publishing, 2016. 
  25. R-documentation on https://www.r-project.org.
  26. Cortes C., Vapnik V. Support-vector networks. Machine Learning. 1995, 20 (3), p.273–297. doi:10.1007/BF00994018.
  27. Cleveland W.S. LOWESS: A program for smoothing scatterplots by robust locally weighted regression. The American Statistician. 1981, 35 (1): 54. doi:10.2307/2683591. JSTOR 2683591.
  28. Data Science and Big Data Analytics: Discovering, Analyzing, Visualizing and Presenting Data. Wiley: EMC Educational Services, 2015. 432 p. 
  29. Machine Learning Course, Stanford University, https://www.coursera.org/learn/machine-learning. 
  30. MATLAB documentation on www.mathworks.com. 
  31. LIGO Gravitational Wave detector experiment on https://losc.ligo.org/events/GW150914/, https://losc.ligo.org/s/events/GW150914/GW150914_tutorial.html.
  32. Scheglov E.G., Arutynov P.A. Non-Archimedean Calculus in the Mathematical Modelling of Digital ICs // Russian Microelectronics. 2002. V. 31. № 6. P. 384–395.
  33. Arutynov P.A. Non-Archimedean Calculi in Theoretical Metrology. Izm. Tekh. (Metrologia). 2001. № 2 P. 3–18.
  34. Varvak M. Ellipsoidal/radial basis functions neural networks enhanced with the Rvachev function method in application problems // Engineering Applications of Artificial Intelligence. 2015. V. 38. P. 111–121. DOI: 10.1016/j.engappai.2014.09.017.
Date of receipt: 18 сентября 2017 г.