A regularized limited memory BFGS method for large-scale unconstrained optimization and its efficient implementations

The limited memory BFGS (L-BFGS) method is one of the popular methods for solving large-scale unconstrained optimization. Since standard L-BFGS uses a line search to guarantee its global convergence, it sometimes requires large number function evaluations. To overcome difficulty, we propose new with certain regularization technique. We show convergence under usual assumptions. In order make more robust and efficient, also extend several techniques such as nonmonotone technique simultaneous use Wolfe search. Finally, present some numerical results test problems in CUTEst, which that proposed terms problems.