A bundle-type method for nonsmooth DC programs

Abstract A bundle method for minimizing the difference of convex (DC) and possibly nonsmooth functions is developed. The may be viewed as an inexact version DC algorithm, where each subproblem solved only approximately by a method. We always terminate after first serious step. This yields descent direction original objective function, it shown that stepsize at least one accepted in this way. Using line search, even larger stepsizes are possible. overall to globally convergent critical points programs. new algorithm tested compared some other solution methods on several examples realistic applications.