350 rub
Journal Neurocomputers №9 for 2013 г.
Article in number:
Associative memory construction based on self-reproducing artificial intelligent neural networks (SAINN)
Authors:
S.D. Ionov - Post-graduate Student, Institute of Mathematics and Mechanics (IMM), Ural Branch of the Russian Academy of Sciences. E-mail: progsdi@gmail.com S.V. Sharf - Institute of Mathematics and Mechanics (IMM), Ural Branch of the Russian Academy of Sciences. E-mail: scharf@imm.uran.ru
Abstract:
In the article «Associative memory construction based on self-reproducing artificial intelligent neural networks (SAINN)» the solution of tasks of processing of the signal flow by creating neural networks that implement associative memory is considered. The article is a description of a specific type of neural networks called the self-reproducing artificial intelligent neural networks (SAINN). The essence of the problem is reduced to the construction of a special neural network for solving a class of problems of signal processing and decision making, using associative memory. This article is divided into 6 parts. In the beginning of the article the authors define the requirements for the solution of the problem and, in particular, define the concept of associative memory and the data connections. Next the overview of SAINN, its main elements and their functionality is presented. The third part provides examples of SAINN construction, solving simple algorithmic and learning problems of neural networks that support the advanced position on the applicability of SAINN in the task. The paper addresses problems of setting up the complex structure of the neural network, and in the fourth part, a possible solution based on specific language describing SAINN - SAINNL - is introduced. In the fifth part based on the proposed main components the detailed description of the associative memory, built using SAINN, is given. In conclusion, the summary on the built SAINN solving the problem of associating signals is made, and it is stated that a similar but more comprehensive solution can be applied for solving the whole class of problems of processing a signal flow.
Pages: 32-44
References

  1. Ionov S.D. Vosproizvodyashhiesya iskusstvenny'e nejronny'e seti: obrabotka i assocziativnoe svyazy'vanie signalov // Sb. tezisov «Sovremenny'e problemy' matematiki». Ekaterinburg: IMM UrO RAN. 2013. S. 313-316.
  2. Siegelmann H.T., Sontag E.D. On the computational power of neural nets// Journal of computer and system sciences. 1995. № 50. S. 132-150.
  3. Widrow B., Hoff M.E. (Jr.) Adaptive switching circuits // IRE WESCON Conv. Rec. 1960. S. 96-104.
  4. Ionov S.D. Mnogovariantnost' primeneniya vosproizvodyashhixsya iskusstvenny'x nejronny'x setej v zadachax raspoznavaniya  // Sb. tezisov XI Vseross. nauch. konf. «Nejrokomp'yutery' i ix primenenie». M.: MGPU. 2013. S. 43-45.
  5. Ionov S.D. Self-reproducing Artificial Intelligent Neural Network Language // https://bitbucket.org/ionsphere/axis4-neural/wiki/Language
  6. Smith R.G. NeuronC User's Manual // http://retina.anatomy.upenn.edu/~rob/ncman2.html
  7. Weitzenfeld A., Arbib M.A., Alexander A. The neural simulation language: a system for brain modeling. SShA, Massachusets: MIT. 2002. 367 s.
  8. Gleeson P., Crook S., Cannon R.C., Hines M.L., Billings G.O. NeuroML: A Language for Describing Data Driven Models of Neurons and Networks with a High Degree of Biological Detail // http://www.ploscompbiol.org/article/info%3Adoi% 2F10.1371%2Fjournal.pcbi.1000815.
  9. Ionov S.D. Associative Memory // https://bitbucket.org/ionsphere/axis4-neural/wiki/Associative%20memory
  10. Compact Flat Sorter Machine Active // http://postalautomation.elsag.it/PDF/schedaCFSM.pdf.
  11. GOST R 51506-99. Konverty' pochtovy'e. Texnicheskie trebovaniya. Metody' kontrolya. M.: Izd-vo standartov. 2000. 20 s.