Iterative algorithms for multiscale state estimation. I: Concepts (Q5956968)

A new class of hierarchical optimization problems with varying discretization schemes which increase the resolution are proposed. The approach is based on Garlerkin discretizations combined with wavelets, which provide good preconditioner and facilitate suitable adaptive refinement strategies. Firstly one describes the model and the source of ill-conditioning, then the corresponding Karush-Kuhn-Tucker system is given in wavelet coordinates, and the optimization so obtained is solved by using an Uzawa algorithm. Then ill-conditioning of the KKK system is considered, and the nested iteration scheme for the resolution of the sequence of systems is outlined. For Part II, see ibid. 111, 501-527 (2001; Zbl 1002.93059).