Finite sample behaviour of an ergodically fast line-search algorithm (Q1303781)

A line search algorithm is presented, based on a study of the ergodic behaviour of the associated dynamic system by a renormalisation of the uncertainty interval at each iteration. It is shown that for functions symmetric with respect to the optimum the algorithm permits to achieve any given precision with significantly less comparisons of function values than the golden section algorithm. For locally smooth functions, numerical simulations are reported that show the robustness of the algorithm with respect to non-symmetry.