Approximation schemes for functional optimization problems (Q1024253)

Approximation schemes and suboptimal solutions for functional optimization problems with admissible solutions dependent on a large number of variables are investigated on the basis of the analysis for a sufficiently large number \(n\) of basis functions in the paper of \textit{V. Kurkova} and \textit{M. Sanguneti} [SIAM J. Optim. 15, 461--487 (2005; Zbl 1074.49008)] in which the critical term is of the form \(n^{- {1 \over 2}}\), multiplied by a certain norm of the optimal solution, called ``variation norm''. In the present paper, the authors show that the upper bounds are of the same order \(n^{-\frac12}\) and the error of the approximate optimization can be obtained for various sets of admissible solutions defined in terms of variation norms with respect to different bases, if one suitably ``matches'' the approximation with the set of admissible solutions. Variation norms are compared for some variable-basis approximation schemes and are estimated in terms of spectral norms. The comparison is given for classical \(L_p\) and Sobolev's spaces with sets of admissible functions having finite variation norms with respect to some bases.