Abstract:
An identification problem is discussed which is described by a Fredholm equation of the first kind. With due regard for a priori data on solution and errors in specifying the right-hand part of the equation, the notion of a minimax solution is introduced and it is shown that in numerical implementation a well developed set of linear algebraical equations should be solved in order to find a minimax solution. A criterion is introduced for selection of dimensionality to this system. To compute a solution, recurrence algorithms analogous to Kalman—Bucy filtering equations are proposed. A numerical example is given.
Fulltext:
PDF file (2217 kB)