A new subspace correction method for nonlinear unconstrained convex optimization problems (Q692728)

The problem of minimizing a real-valued differentiable function over a finite dimensional vector space is studied. Following the framework of \textit{X.-C. Tai} and \textit{J. Xu} [Math. Comput. 71, No. 237, 105--124 (2002; Zbl 0985.65065)], the authors construct a new parallel subspace correction algorithm. The subspace problem, based on the multigrid approach for discretized optimization problems proposed by \textit{S. G. Nash} [Optim. Methods Softw. 14, No. 1--2, 99--116 (2000; Zbl 0988.90040)], is more close to the original problem, which makes the subspace correction more accurate. The convergence rate of the algorithm is given and numerical results for nonlinear partial differential equations and a minimal surface problem are shown.