Русская версия
ISSN 1817-2172, рег. Эл. № ФС77-39410, ВАК

Differential Equations and Control Processes
(Differencialnie Uravnenia i Protsesy Upravlenia)

About History Editorial Page Addresses Scope Editorial Staff Submission Review For Authors Publication Ethics Issues Русская версия

Visualization of Compositions of Multidimensional Objects with Local Descriptions in Metric Machine Learning Algorithms

Author(s):

Viacheslav Anatolievich Diuk

Doctor of Technical Sciences,
Principal Researcher of the Institute for Transport
Problems of the Russian Academy of Sciences (IPT RAS)

v_duke@mail.ru

Abstract:

In various fields, artificial intelligence (AI) models are increasingly used for decision-making based on machine learning. In metric machine learning methods, objects are treated as precedents, and only one operation is used: determining the similarity (difference) between these precedents and an unknown object. The main limitation of existing metric methods is related to representing a common feature space for all objects and, consequently, a single measure for measuring distances between objects. This limitation is overcome by constructing a unique local feature space for each object and finding individual measures that determine the hierarchy of its similarity to other objects, relevant to the given context. The article discusses the problem of analyzing sets of objects with local descriptions and proposes a solution using d(S)-metrics, which reflect differences in distance series between the same objects but in different local spaces. Introducing d(S)-metrics allows for further visual analysis of compositions of objects with local descriptions using multidimensional metric scaling. The article provides a practical example of such analysis in the task of recognizing vehicle types based on geometric features of their silhouettes.

Keywords

References:

  1. Fix E., Hodges J. L. Discriminatory analysis: nonparametric discrimination: consistency properties. - Rep. N 4. - USAF school of Aviation Medicine. - Texas. - February 1951. - Project 21-49-004. - Contract AF-41-(128)-31
  2. Zhuravlev Yu. I., Nikiforov V. V. Recognition algorithms based on calculation of estimates // Cybernetics. - 1971. - 1-11 p
  3. Zagoruiko N. G. , Borisova I. A. , Dyubanov V. V. , Kutnenko O. A. Methods of Recognition Based on the Function of Rival Similarity//Pattern Recognition and Image Analysis, 2008, Vol. 18. No. 1. pp. 1-6
  4. Vapnik, Vladimir N. The nature of statistical learning theory. : Springer-Verlag New York, Inc., 1995
  5. Cover T., Hart P. Nearest neighbor pattern classification//IEEE Trans. Inform. Theory, v. IT 13, 1967. P. 21-27
  6. Duda R., Hart P. Pattern recognition and scene analysis. - M. : Mir, 1976. 509 p
  7. Duke V. A. Experimental study of the response of machine learning algorithms to data labeling errors // Differential equations and control processes. Electronic journal. - http://diffjournal.spbu.ru/ , 2022. No. 3. pp. 59-72
  8. Boulding K. E. General Systems Theory - The Skeleton of Science // Management Science, 2, 1956
  9. Gick J., van. Applied general systems theory. - M. : Mir, 1981
  10. Duke V. A. Computer psychodiagnostics. - St. Petersburg: “Brotherhood”, 1994. - 364 p
  11. Townsend K., Focht D. Design and software implementation of expert systems on personal computers. - M. : Finance and Statistics, 1990
  12. Krylov V. Yu. Method of multidimensional geometrization of psychological data. Systematic approach in mathematical psychology // The principle of systematicity in psychological research. - M. : Nauka, 1990. - P. 33-48
  13. Duke V. A., Malygin I. G., Pritsker V. I. Recognition of vehicles by silhouettes - a three-stage method of machine learning in technical vision systems // Marine intelligent technologies. - 2022. - No. 2-1(56). - pp. 162-167
  14. Dyuk, V. A. Context-dependent local metrics and a geometrical approach to the problem of knowledge formation / V. A. Dyuk // Izvestiya Rossiiskoi Akademii Nauk. Teoriya i Systemy Upravleniya. - 1996. - No. 5. - P. 36-44
  15. Handbook of Applied Statistics. In 2 volumes. T. 2 // Ed. E. Lloyd, W. Lederman, S. A. Ayvazyan, Yu. N. Tyurin. - M. : Finance and Statistics. - 1990
  16. Saito T. The problem of the additive Constante and eigenvalues ​ ​ in metric multidimensional scaling // Psychometrika, v. 43, No. 2, 1978
  17. http://www.ics.uci.edu/~mlearn/MLRepository.html (accessed 10. 06. 2024)
  18. https://archive.ics.uci.edu/ml/datasets/Statlog+(Vehicle+Silhouettes ). (accessed 10. 06. 2024)
  19. Duke V. A., Malygin I. G. Comparison of algorithms for recognizing types of vehicles based on the parameters of their silhouettes // Marine intelligent technologies. - 2018. - No. 4-4(42). - pp. 197-201
  20. Duke V. A., Malygin I. G., Pritsker V. I. Recognition of vehicles by silhouettes - a three-stage method of machine learning in technical vision systems // Marine intelligent technologies. - 2022. - No. 2-1(56). - pp. 162-167
  21. Swamidass S. J., Azencott C., Daily K., Baldi P. A CROC stronger than ROC: measuring, visualizing and optimizing early retrieval // Bioinformatics (2010) 26 (10): 1348-1356
  22. Torgerson W. S. Multidimensional Scaling. Theory and Method // Psychometrika, v. 17, No. 4, 1952
  23. Davison M. Multidimensional scaling: Methods for visual presentation of data. - M. : Finance and Statistics, 1988

Full text (pdf)