2. Home
  3. ARCHIVE
  4. 2021
  5. VOL 43, NO 3 (2021)
  6. pp. 36-46

SYSTOLIC ARCHITECTURE OF MATRIX PROCESSOR FOR CLASSIFIER OF OBJECTS

T.B. Martyniuk, L.V. Krupelnytskyi, M.V. Mykytiuk, M.O. Zaitsev

Èlektron. model. 2021, 43(3):36-46
https://doi.org/10.15407/emodel.43.03.036

ABSTRACT

One of the known methods of object classification is considered, in which the criterion of classification by the maximum of discriminant functions is realized. This method has found effective application as a classical computational model, in particular in the medical diagnosis of diseases. The process of classification by this method can be implemented as spatially distributed processing on columns and rows of the matrix, which can be described as regular iterative algorithms. This allows you to display them on a two-dimensional systolic array of the matrix processor as part of the classifier of objects with subsequent placement in the FPGA. The proposed matrix processor works in two modes and has a number of specific properties, such as performing the decrement operation simultaneously for all elements in each column of the processor matrix, as well as the use of zero signal (zero) of elements in each row and each matrix column as a result of processing of discriminant functions and for synchronization of the process. In the future, based on the results of processing in the matrix processor, the output signals of the classifier are formed with the definition of a specific class of objects.

KEYWORDS

systolic architecture, discriminant function, classifier of objects.

REFERENCES

  1. Meribout, M. and Firadus, A. (2016), “A new systolic multiprocessor architecture for real-time soft tomography algorithms”, Parallel Computing, Vol. 52, pp. 144-155, available at: https:// doi.org/10.1016/j.parco.2016.01.002.
    https://doi.org/10.1016/j.parco.2016.01.002
  2. Bagavathi, C. and Saraniya O. (2019), “Evolutionary Mapping Techniques for Systolic Computing System. Deep Learning and Parallel Computing Environment for Bioengineering Systems”, Academic Press, pp. 207-223, available at: https://doi.org/10.1016/B978-0-12-816718-2.00020-8.
  3. Korchenko, A.H., Kynzeryaviy, V.N., Hnatyuk, S.A. and Panasyuk, A.L. (2010), Sistolicheskiy kriptoprotsessor [Systolic Crypto Processor], Mizhnarodna naukovo-praktychna konferentsiya [International scientific and practical conference], Informatsiyni tekhnolohiyi ta kompyuterna inzheneriya [Information technology and computer engineering], Vinnytsya, VNTU, May 19-21, 2010, pp. 187–188.
  4. Martyniuk T.B., Kozhemiako A.V. and Khomyuk V.V. (2009), “Systolic Array Models for Processing Vector Data Using Difference Slices”, Upravlyayushchiye sistemy i mashiny, Vol. 5, pp. 46-55.
  5. Yadzhak, M.S. (2011), “Features of realization of artificial neural networks of one type on quasi systolic computing structures”, Computational Intelligence (Results, Problems and Perspectives), Proceedings of the First International Conference, McLaut, Cherkasy, Ukraine, pp. 134-135.
  6. Kun, S. (1991), Matrychnye protsessory na SBIS [VLSI array processor], Myr, Moscow, USSR.
  7. Kanevskiy Yu.S. (1991), Sistolicheskye protsessory [Systolic processors], Tekhnika, Kiev, USSR.
  8. Tymchenko, L.I., Martyniuk, T.B. and Zahoruyko, L.V. (1998), “An approach to organizing a multilayered systolic computation scheme”, Elektronnoye modelirovaniye, Vol. 20, № 5, pp. 33-42.
  9. Yadzhak, M.S. (2005), “Modeling of neural networks with projective and lateral connections on quasi-systolic structures”, Vidbir i obrobka informatsiyi, Vol. 23, № 99, pp. 122-127.
  10. Martyniuk, T.B., Kozhemiako, A.V., Krupelnytskyi, L.V., Perebeynis, O.M. and Bezkrevnyy, O.S. (2016), “Implementation models of a matrix processor for a biomedical data classifier”, Informatsiyni tekhnolohiyi ta kompyuterna inzheneriya, Vol. 2, № 36, pp. 43-51.
  11. Martyniuk, T.B., Buda, A.H., Khomyuk, V.V., Kozhemiako, A.V. and Kupershtein L.M. (2010), “Biomedical Signal Classifier”, Iskusstvennyy intellekt, Vol. 3, pp. 88-95.
  12. Ranhayyan, R.M. (2007), Analiz biomeditsinskikh signalov, Prakticheskiy podkhod [Analysis of biomedical signals. A hands-on approach], FIZMATLIT, Moscow, Russia.
  13. Yunkerov, V.Y. and Grigoriev, S.H. (2002), Matematiko-statisticheskaya obrabotka dannykh meditsinskikh issledovaniy [Mathematical-statistic processing of medical research data], VMedA, Saint Petersburg, Russia.
  14. Diskriminantnyy analiz [Discriminant analysis], available at: http://www.statsoft.ru/home/textbook/modules/stdiscan.html.
  15. Martyniuk, T.B., Kozhemiako, A.V. and Kupershtein, L.M. (2018), Aspekty raznostno-srezovoy obrabotki dannykh v neyrostrukturakh [Aspects of difference-slice data proces­sing in neurostructures], LAMBERT Academic Publishing RU.

Full text: PDF