Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.cosmos.ru/seminar/199912/e_abstract.htm
Дата изменения: Thu Dec 23 16:06:51 1999 Дата индексирования: Tue Oct 2 10:02:50 2012 Кодировка: Поисковые слова: п п п п п р п |
Main page
Reports Discussion |
A mathematical analogy between a L-and MV-optimal experiments design and an ideal linear correction of system parameters is established, Optimality conditions are derived on the basis of this analogy and effective algorithms for solving these problems are designed. Mentioned algorithms are reduced to solutions of some
generalized linear programming problems and its complexity not exceeds the usual linear programming problems complexity with m+1 ariables and m constraints - equalities, where m is the dimension of a estimated parameters vector in a linear regression model.
The performance of these algorithms is demonstrated by
solving a problem of a measurements optimal distribution for a
polynomial regression, for which an analytical solution of the
MV-optimal problem is found.
The E-optimal problem is reduced to the generalized linear programming
problem with M+1 by constraints, where M=m (m+1) /2. The difference derived
solutions from the usual linear programming problem is that the optimality
condition check is reduced to finding the eigenvalue of a MxM-matrix.
The report contains outcomes of researches spent by Voiskovskii M.
and Soloviov V. together with the author. |