Affine scaling interior Levenberg-Marquardt method for bound-constrained semismooth equations under local error bound conditions (Q935779)

An affine scaling Levenberg-Marquardt method with nonmonotonic interior backtracking line search technique for solving bound-constrained semismooth equations under local error bound conditions is presented. Other possible norms (inner products) which may significantly change the computational behaviour (but usually not the convergence rate), effectively providing a preconditioning, are not mentioned. The global convergence results are developed for a semi\-smooth Levenberg-Marquardt method with a backtracking line search technique projecting trial steps onto the feasible interior set. Close to the solution set the algorithm is shown to converge locally Q-superlinearly under some quite weak conditions: The line search as presented is very simple, and maybe computationally very expensive. Overall this is a super-linearly convergent algorithm for non-smooth systems of equations with constraints.