Synthesis of Predicting Filters by Methods
of Pattern Recognition Learning

V. I. Vassilyev* and Yu. I. Gorelov**

* Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine,
pr. Akademika Glushkova, Kiev, 252207 Ukraine

** Velikie Luki State Agricultural Academy, Velikie Luki, Pskovskaya oblast, Russia

Abstract—Problems of synthesizing predicting filters for dynamic systems whose behavior can be described
by evolutionary operator equations are discussed. For a solution to this problem to be well-defined, a class of
regular prediction problems is introduced. The question whether or not a particular problem belongs to this class
can be answered by checking certain conditions on the basis of qualitative or quantitative a priori information.
It is suggested that the synthesis of a prediction filter be transformed into the problem of search for a recogniz
ing algorithm by methods of pattern recognition learning on the basis of initial data transformed in a certain
way. The quality of approximating and extrapolating the initial data by the methods suggested in this paper is
estimated. Some common special cases of solving the problem of reconstructing dynamic dependences with
the purpose of applying them to prediction are described.

Pleiades Publishing home page | journal home page | top

If you have any problems with this server, contact webmaster.