Adaptive frame methods for nonlinear variational problems (Q2480889)

An adaptive numerical solver for certain nonlinear variational problems is developed. The following abstract problem is studied: Find \(u\in V\) such that \(Au+ A_1(u, u)= l\), where \(V\) is a Hilbert space and \(l\) is a functional on \(V\). A is assumed to be an elliptic operator. The discretization of the variational problem is done by suitable frame decomposition (stable, redundant and nonorthogonal expansions). The derived discrete fixed point iteration can be realized efficiently using key routines to approximate the solution of the elliptic problem adaptively by means of Richardson type iteration. The authors investigate the convergence, stability, and quasi-optimal complexity of the algorithm. A construction of suitable multiscale divergence-free frames on a bounded domain is presented. These aggregate divergence-free wavelet frames preserve good properties of standard wavelet bases. The mentioned frames are usually easier to construct than Riesz bases.